Ai’s 3 parts: ( Machine learning ML, Deep Learning DL, & Data Science DS)

In this section we will talk about arguably the most complicated part of the robotics puzzle… AI.

Simply put, AI is the main overall category and within it there is the subcategory called machine learning and then below that is another subcategory called deep learning and to tie it all together is the area called data science.

To really understand and be a programmer in this area, you must learn dozens of terms in that areas vocabulary and then get a solid grounding in advanced math areas like calculus and linear algebra and then know how to deal with things like matrices, vectors, activation functions and derivatives and then know how to program with languages like Python.

Some say that the quickest or easiest way to program in AI is to use higher level environments like Tensorflow, Pytorch or Torch vision and take advantage of pretrained routines that they have.

You could try to get a fast start with AI by getting your webcam ready and using the OpenCV website where you could download the opensource libraries and learn how to use some prewritten computer vision routines.

OpenCV or (Open source computer vision library) is a term thrown around a lot on AI discussion forums and it is an amazing resource. It is basically an open source library of computer vision and machine learning software that makes it easier to get into this topic.

At this time in 2021 there are over 2,500 algorthms there and most of them have to do with facial recognition and object recognition. You will see your face on your computer screen and you would train it or perhaps its already trained to have dots on your face or hands and then you would modify the code on your screen…probably Python.

Certainly in the ideal world, we would all be born programmers and be able to learn how to code in any language and have the necessary discipline and brain focus to think out the programming problem logically and then code and degug the dozens or possibly thousands and thousands of lines of code in a big project or at least know how to maneouver your code with existing pre written libraries.

So lets not be too high and mighty and criticize all the code pasters out there who are at least making an effort which may one day lead them to create original code.

But yes, thankfully you can either let other people program AI for you or you can just strive to be a code paster or luckily you can be somewhere in the middle and be thankful that alot of this is already mostly done for us and you can download and use the following types of tools that make it accessible to the semi skilled user:

  • Tensorflow
  • PyTorch
  • Theano
  • Keras

But for now….lets give some definitions like what is AI versus machine and deep learning and data science?

Well simply put…AI is a branch of computer science which seeks to develop intelligent computer machines that can think and act like humans.

AI is something that can be used to build a self driving car or it can be used to create a human like robot so that one day soon you may think that you are having a conversation with a real human in front of you but in fact it is a robot with very good AI. Or AI doesn’t even have to be in something physical it can appear as a chat bot on your computer screen.

AI can be installed on a website in the form of an AI chatbot…so that when you ask the person on the other end of the website a question you may be getting answers that sound exactly like they are coming from a live person when in fact they are only coming from a very good AI application.

Microsoft had a failed chatbot experiment in 2016 on Twitter when it put out a chatbot named Tay, to learn from others on the site but had to take it down after only about 16 hours.

Now more recently, for some laughs you can watch some product training videos on how to create avatar chatbots with an amazing company called

These videos feature Jason Gilbert a creator on Cocohub with his highly entertaining and unique style along with some very Jack Black’esque flourishes.

Eventually the Ai experts will be able to synch up an artificially created face on a website and turn it into a naturally moving live video image of some person by using some software technology called Gan.

One day soon you could be having a video support call with what you think is a real person talking to you on the website with his or her lips moving and eyebrows that are going up and down and the AI person might be laughing at the jokes you are making and telling you some in return that are funny.

But that could be an example of a virtual video AI chat bot.

Is this future use of AI a good thing? Well thats a good question…but we are here on toystowork website to explain it ….not so much to judge it.

Now AI is all about computer programming and advanced mathematics for the most part.

It has the sub categories of machine learning and deep learning and data science so lets define those terms.

Its all about data and recording mistakes that the program makes and correcting them over and over and over again until it has a huge storage area full of past experiences that will make the AI device do things without having a human telling it what to do or where to go.

Now the term machine learning is like what it sounds like….its where computer machines will learn over time to get better at whatever they are doing.

It is a collection of algorithms or computer programs that have been written to work with massive amounts of data and information and to organize it in such a way that it can learn from it more every time it accesses it.

There is a great Youtube machine learning series by a young person named “Jabrils” who gives some great tips on how to learn machine learning. Its his struggle as a autodidact who learns well on his own as he tries to master the basics.

You will hear about different was to work with macine learning like supervised versus unsupervised machine learning.

This is why we have seen companies like Google construct buildings upon buildings housing rooms and rooms of computer servers which have probably….been recording almost every keystroke and search question that you have ever typed on a webpage since the company began.

This information is housed and can be processed which is called data mining which means that the useful information can be scraped out of the useless bits and turned into something else that is useful to somebody, usually someone who pays for this new gathered information.

Often the useful information tells an advertiser how to better sell us more and more products.

There are 3 main ways to process data using deep learning.

  • clustering
  • classification
  • regression

Now deep learning or deep multi neural learning is a category below machine learning and is part of the big boss of Artificial Intelligence.

So while machine learning is about getting the computers to learn more on their own without help from a human, deep learning goes a bit further and tries to learn things more by following certain structures that are based on the human brain.

Deep learning really wants to learn how to do things more like a how a human brain would actually do things and here you will find terms like ANN and CNN and RNN.

This is sort of the area where Elon Musk has been talking about in his television and web interviews when he talks about AI and his neural networks and his plans to make a physical link from a persons brain to a computer chip.

Then we have the last category called data science which is all about using advanced mathematics like linear algebra, probability and statistics to make it all make sense and put the whole AI package together.

Now data science is relatively new and is related to the term data mining which was big in around 1996.

Computer science was combined with data mining and had a baby…data science was born. This was around the time of Web 2.0 when all the various websites were sharing data and picking things from each other.

To get involved in data science you have a handful of areas or systems to work with an it all involves working with data and its all about solving real world problems for the organization using the mountains of data that the company has been gathering and probably not using to its advantage.

Many refer to the top users of what was called Big data are companies like Facebook, Google, Amazon, Apple etc.

You will be involved in the following systems.

  • gather the data in some way such as off the internet and will use some web scraping tools such as Beautiful soup, Scrapy and URLLib.
  • pick your language like Python or R or Java
  • get an ide (integrated development environment) to write your code . Pycharm or Jupyter
  • get involved in math like statistics, linear algebra and diferential calculus
  • use data visualization

At some point you will get involved with AWS or Microsofts Azure.

The best way to learn about all of these things is to take a free or paid course online or watch the hundreds of youtube videos on these topics.

Start by learning these terms and many more and then download a few of the tools like Tensorflow and Pytorch and Keras and Theana and Microsofts Azure etc etc.

You could buy a $200 Nvidia Jetson Nano developer kit computer and start actually creating some AI applications.

You could get more involved in the math and start learning about basic linear algebra and differential calculus and matrices mathematics. Then look at the compicated seeming circles upon circles which are trying to outline various neural networks etc.

Be overwhelmed but don’t give up because more and more tools are arriving every day so that you can almost just cut and past certain bits of code in to do some amount of AI that you may need or want to do.

If you already have a PHD degree in computer science or physics with a major in advanced mathematics and programming in languages like Python and C++ and Lisp etc then you are at a definite advantage.

But do look up some of the videos with Andrew Ng and Elon Musk where they give their explanations for where AI may be going.

There are basically the two camps with Elon saying that none of the smart people in the AI sector know what they don’t know and that they should be afraid of AI…or at least a lot more concerned.

Elon believes that there will be such incremental or that there have been such exponential advances in AI in a short few years that we will rapidly see the subject of AI and superintelligence overtake us.

He talks about how the AI programs like Alpha Go were able to learn to beat the masters so quick that it was hard to believe and that a better AI program after that was able to learn to play just about any game so fast that everybodys heads were spinning.

So where we are going with AI is up for discussion.

For now we will try to get beginners to understand some of the basic concepts and then let each person decide if they are up for the mathematical challenge which is a big part of learning machine learning and deep learning…areas like

  • linear algebra
  • probability
  • calculus
  • optimization

Some other terms and concepts that you will come across are singularity and the father of all things AI…the Turing test.

We will cover some of the areas in this whole topic of AI like vision systems, object recognition, GAN, Tensorfo, Python etc

Deep learning at places like Carnegie Melon, Stanford, Google, Facebook..

Famous movies or documentaries that reference AI like “Game Over” which profiled the chess game between Kasperov and the IBM Deep Blue computer where the computer won.

How the Watson program once played Jeopardy.

We will ask if the computer will transcend the human mind….and here we talk about singularity.

Several important figures in the AI community will be discussed including Minsky, Kurzweil, Chomsky on linguistics.

Some interesting things are being done with what they call BWI Building Wide Intelligence bots at the University of Texas involving longe range ai etc.

Hanson Robotics is doing some interesting things with their Sophia robot and encouraging a world wide movement through a foundation to spread AI and improve on how Sophia operates. We need to read more about it but there is a foundation and a set of goals and mention of the word DAO and a website Singularitynet. We will look into it more in July 2021.

They have some code on the Github site and there are development tools that are available and much more.

We will have to see if other major robotics companies are working in cooperation with this idea or if every major company will be going at it on their own.

There are questions of how far ai can go and will we see a super ai capable of much more with the theory of embodied cognition.

We will look more into areas like

Microsofts Azure Cognitive services in which 5 areas are covered including vision and speech.

Under Open AI which has both a for profit and a non profit side you can learn about GPT 2 and Gym.

Elon Musk was one of the original founders of Open AI and at some point Microsoft invested $1 billion in it.

Googles Coral edge TPU is an AI initiative which apparently is an attempt to be able to have ai without using the cloud.

” Gazebo is an open source 3d robotics simulator.

and Opencl point cloud libraries,

Tensorflow is a Google ai product which is an open source deep learning framework. It is a library that can work with Python.

DeepMind Technologies from Alphabet based in England have their neural networks projects that can play video games and are known for their work with a Neural Turing machine. They also showed off a Go game playing form of AI in 2016 and have had several revisions to it since.

They use a different type of AI and is not preprogrammed but learns from experience and apparently deals with raw pixels and uses what they call deep learning on a convolutional neural network.

, computer vision with (shore and Visage), MITs chatscript,

There is a $400US product from that allows you to experiment with a form of neural networks and your brain.

, neuro catch,

ai controlled prosthetics with electro myography or ultrasound…Mark Tildens NV breadboard and biomech motor nacelles and nervous nets.

We can talk about working with vision systems by starting with toys like the Xbox Kinect camera and conmpare it to the Occuls Rift for affordable vision system entry, speech recognition and word spotting function used in Aldebran/Softtbanks Nao robot.

Now lets talk a bit about AI and such technologies and how they have been portrayed in the movies in the last 25 years or so.

Some movies with the AI theme like….”A.I” in which an employee at an AI corporation is given a chance to live with an artificial boy just after their own child passed away.

Movies like “Her” where the main character falls in love with an operating system.

Others like “Lars and the Real Doll”.

“SIMONE” with Al Pacino.

Others of course like The Terminator, the Matrix, Lawnmower man, 2001 Space Odysey…and many others.