Machine learning

AI and the Terminator era begins. Machine learning takes over.

This AI apocalypse is not so scary for me because computers don’t have two important components: initiative and feelings.
We may even give them one in the future but they will need both ingredients to make the seed of revolution.

The danger these days is not with the machines but with people that control and operate them. Is like “it’s not the gun but the owner that pushes the trigger” that is the problem.

Lets get back in time a little and look at how we thought the future will look like.

In the 50-60s movies were showing year 2000 as something futuristic, flying cars, strange looking robots, aliens and megacities with 1000 level heights.

That was believed to be if not true but at least partially true, it was a hope for the future.

Now if you come back to present day year 2020, which is way beyond the scifi books and presictions. Just look around.

How many things were disrupted, how close we are to that dream?

Visually we are not much different than 70 years ago.

Lets count on a few of the biggest changes:

  • Medicine and body reparation, change
  • Commonality of flying 
  • Internet and communications(big data)
  • Education
  • Life expectancy

Last point is actually the power that fuels the rest.

So in reality a Terminator style killing robot with beyond human level of perception and will is very very far from us. ( maybe 1000 years from now)

People tend to be scared of things they don’t understand and AI is no different.

I will tell you eight now what machine learning and AI really are: analytics and probabilities.

ML borrowed our human brain neural network’s model and used it to make simulations on huge amounts of data. This is in fact how we learn.

It is the opposite of programming when you have to think about all possible scenarios and write code around it to get results. In ML you just give the huge amount of data to the computer and it makes sense of it by itself.

How machine learning works.

I will try to explain as best as possible how ML works in real life.

You must have a computer, Python installed and DATA. ( I am writing DATA with capital letters because you need a lot of it).

What kind of data Vlad? Well I will give you some examples:

  • Customer reviews on a websites
  • Client spending (purchase data)
  • Pattern identification in images
  • Traffic data from the streets
  • Real estate data
  • Etc. etc

Now that you have the data in excel format or whatever you need to make sense of it because some of it is not helpful. For example ip addresses list from locations, or middle name of a person and other things that are not related to what you require from the data.

You can approach the problem in two ways. Once can be that you already know what you want from the data. And another that you have no idea, but you want to find some gems. Some correlation that can is important.

In any case you have to split the data in 3 parts.

  • Training Set
  • Test Set
  • Validation Set

The percentage is debatable but the bigger the training set is the more accurate the results, the more data the better prediction.

So now you built the model in Python (or R) and are satisfied with the prediction results you can use it forever on new data.

Ok but how will this change my life?

Let me explain please.

The more data we collect the better models can be created and more accurate “smart data” is generated. This smart data is optimizing everything around us: selfdriving cars, smartwatches, train schedules, airport control, ads on websites, Spotify recommendations etc etc.

Everything around you got smarter and more accurate so it is influencing you in a deeper way in the long run.

If for example the cookies in your browser are fed to a ML system and that system detects that you fit in a particular category of people you may and will be targeted with such “personal” ads that you will think it is magic.

It may influence not just what you buy but also who you vote for, what school your kids will go to and the type of food you will eat.

You may like it or hate it , it is the future we are heading.

Also lately it is common not to own anything. Before we were buying a movie on a DVD for example. Now the trends showed that owning stuff is subject to piracy so subscription based systems are put in place.

The house you “own” is basically rented from the bank, the internet connection is subscription, the electricity is subscription and the Netflix is guess what?.. subscription.

It gives you flexibility but it also sets you up for a neverending loop of need towards this companies that own your personal details and more importantly your preferences.

You may think your home address or phone number or even bank account is “personal” and “private” stuff that you are afraid to share but the most important aspect of you is your “thoughts” and intentions.

That is what ML is used for right now in ads for example. They know what you’ll buy before you even know it so they feed you just things you want to see.

The choices you’re making online, the pictures you like the comments you put, the things you google are FOREVER stored no matter if you delete them.

So this data is used and sold online by this big companies.

Machine learning algorithms

Lets describe a little what we know about the machine learning algorithms. The field is in constant innovation so it is just what we know until now and what is available publicly.

  • Supervised Learning

How is done: This algorithm consist of a target / outcome variable (or dependent variable) which is to be predicted from a given set of predictors (independent variables). Using these set of variables, we generate a function that map inputs to desired outputs. The training process continues until the model achieves a desired level of accuracy on the training data. Examples of Supervised Learning: Regression, Decision Tree, Random Forest, KNN, Logistic Regression etc.

  • Unsupervised Learning

How is done: In this algorithm, we do not have any target or outcome variable to predict / estimate. It is used for clustering population in different groups, which is widely used for segmenting customers in different groups for specific intervention. Examples of Unsupervised Learning: Apriori algorithm, K-means.

  • Reinforcement Learning

How is done: Using this algorithm, the machine is trained to make specific decisions. It works this way: the machine is exposed to an environment where it trains itself continually using trial and error. This machine learns from past experience and tries to capture the best possible knowledge to make accurate business decisions. Example of Reinforcement Learning: Markov Decision Process

Best Python Libraries for Machine Learning

I will go into details later on each of them (maybe separate post? ) but generally the ML in Python can be done using the below list of libraries (free of charge)

The most important is Scikit-learn library for sure.

  • Numpy
  • Scipy
  • Scikit-learn
  • Theano
  • TensorFlow
  • Keras
  • PyTorch
  • Pandas
  • Matplotlib

Machine Learning engineer salary these days is $114121 in average so it may be motivating enough to search for a job in the field 🙂

What about the good Vlad?

Yes yes, I know what you are thinking: how horrible this ML is and look what people use it for, making money.

But that is not the whole story.

Pharma companies use it for molecule modeling and finding cures for the worst types of diseases with the help of big data and analytics.

IBM Watson helps medicine as well by customized treatments given to patients after entire body profile and scans are feed to the “machine”.

In real estate correct prices are calculated/estimated for people by zone.

Trips are scheduled based on the customer preferences and needs.

So I leave you with a question: knowing what we know now about machine learning and it’s power, how will you let it affect your life?

Cheers and thank you for reading TheLongWalks




Please enter your comment!
Please enter your name here