Is Machine Learning turning us into Remora fish?

Events that led to it

Similar to AI, Machine Learning (ML) has been around for some decades. This a brief description of some events that have helped the creation and development of this field.

If you take a look at the events, per decades, you see that 50% happened in the last 12 years! Basically in the first 50 years of existence Machine Learning had two major problems: computational capability and available data. In the last decade, we’ve seen a “boom” in available information (photos, documents, messages, etc), as well as an increase in both computational power and processing speed, much do to gamers demands (thank you Gaming industry!). Moreover, it is curious to see that the big companies such as Facebook, Amazon, Microsoft, and Google are responsible for most of these latter events which means they’re responsible for taking ML to the next level.

Interesting fact is that IBM seems to love building machines that defeat humans in their leisure hobbies. Bad, bad IBM!

Data, Data is everywhere!

Everywhere you go there is Data! You probably don’t know, but there is a term for it : Data exhaust. Basically, every day you go to the Internet and every time you access the web you leave a trail of your activities( Facebook activity, text messages, GPS location, etc). Every time you touch your phone new data is created, want it or not. The only possible way you have to avoid this is to live in a desert island like Chuck Noland (interpreted by Tom Hanks) and Wilson the Volleyball.

Data is growing exponentially, however this has only been the case in the last couple of years.

Since the beginning of Mankind until 2005 humans had created 130 ExaBytes of Data.

“But what’s an ExaByte?” — asks the reader. In 2018 we’re already familiar with what’s a TeraByte (TB). One TB is equal to 1000 GB. So, if a movie usually occupies 1GB in your computer’s memory than you can have almost 1000 movies in a 1TB external disk. That is a lot, right?

Now, what if I tell you that 1 ExaBytes (EB) of data is equal to 1 000 000 TB? Imagine how much information you have to generate to fill up an external disk with that size! That’s what mankind has been doing since the dawn of times, all of the books written, all sang songs, all words spoken, etc.

Mind blowing right?

According to IDC’s Digital Universe Study, in 2010 that number was already around 1 200 EB!! Hence, in only five years, we created much more data than ever before. Later, in 2015 the number grew to 7 900 EB! In 2020 this number is estimated to be around 40 9000 EB of data!

Source

But how does it work?

By now we can say that Machine Learning is the semi-automated extraction of knowledge from data. It is a mix of Statistics and Computer Science, that enable computers to do certain tasks. But how the hell does it work?

Machine Learning can be divided into three components :

  • Data – every ML problem starts with a question that might be answerable with data;
  • Pattern insight – A computer provides insight on the data;
  • Semi-automated – Requires many smart decisions made by a human.

In the end, it is like going back to Agriculture, but a little bit more technological. In order to get a computer to programme itself we have to have the following components:

  • Seeds = Algorithms
  • Nutrients = Data
  • Gardener = You
  • Plants = Programs

“Machine Learning: What it is and Why it Matters” by Priyadharshini

  • Understanding with examples

Suppose you have all registers of sold coffees from your favorite Caffe store, for each hour of the day for the whole year. We can plot this data to look for some type of pattern. Take a look at the image below.

Source

We can clearly see that the highest number of coffees sold is between 8 to 10 a.m. If the Caffe owner asked us what is the period of the day that he needs to be the sharpest to serve a high number of coffees we would point to that part of the day, correct? With ML, a machine is able to learn this pattern just as easily as we did and then make predictions based on what it has learned, just as we humans.

Now, this example is very simple even for us mere mortals. However, imagine we add more data to this graph such as “Type of Coffee”, “Price of Coffee”, “Consumer’s Age”, “Type of Payment”, etc. We can end up with a multidimensional dataset(hundreds or thousands of features) instead of this very simple and easy two-dimension example.

For us it is impossible to analyze so much information at the same time and look for patterns in it. Nevertheless, with the most recent computers, which provide high computational power, it is a walk in the park. Again, once the computer has learned the patterns from the data it can make predictions that humans cannot come close to.

Let’s take another example, just to settle this idea. Imagine you need a computer that can tell you the difference between the image of a dog and another of a cat. We can start by feeding it images and label those same pictures indicating to the computer which one is a dog and which one is a cat (Supervised Learning).

In order to learn with the given data, our computer will look for statistical patterns that might help it differ cats from dogs in the future. It might figure out that cats have shorter noses and that dogs come in a larger variety of sizes. In the end, it is the computer not the developer that figures out those patterns which will establish the algorithm by which future data will be sorted. As you add more data to the system its performance will improve.

Benefits

Here are some benefits that daily users can take from using ML algorithms and applying them to their routine tasks.

“Machine Learning: What it is and Why it Matters” by Priyadharshini

Applications of Machine Learning

  • Image Recognition

Source

  • Facial Recognition
Resultado de imagem para machine learning facial recognition

Source

  • Self-driving cars
Resultado de imagem para machine learning self driving cars

Source

  • Text and speech recognition
  • Online viewing recommendation (Netflix, Spotify, Amazon, YouTube)
  • Spam filter
  • Credit Card fraud detection
  • Medical Diagnosis
  • Stock Market Analysis
  • Sport Match Reports

Conclusion

I guess it is safe to say that Machine Learning is here to stay! This type of Artificial Intelligence allows us to process great amounts of data in a shorter period of time while at the same time empowering us with better and more accurate decision making capabilities. Moreover, these systems will only get faster, better and more inexpensive as technology evolves.

read original article at https://medium.com/diogo-menezes-borges/is-machine-learning-turning-us-into-remora-fish-d3effa0ea56a?source=rss——artificial_intelligence-5