During the last couple of years, the interest for artificial intelligence or AI has increased. In the podcast Concerning AI hosted by Ted Savarta, business growth advisor and Brandon Sanders, PhD in Computer Science they talk about the recent years as being an AI spring because of the increasing interest. Companies such as Google and Amazon have both released consumer electronics with intelligent assistants and both iPhones and Android provide assistants using AI. But AI is not a new invention. Standford University publishes online courses and amongst them an AI or more specifically a Machine Learning or ML course. This course has been available for around 10 years and includes examples of image recognition and autonomous cars driving on public roads.

So how come that AI just recently became a buzzword for most of the American IT companies. Moore’s Law is one of the reason and as Intel describes the law impacted on three different levels: The law has an impact on economics; since Moore’s law prescribes that transistors will fit into gradually fit into smaller places increasing the performance and with this increase, the product becomes more effective allowing for cheaper computing. This leads to the technology level which is impacted because cheaper computing gives the opportunity to make more complicated systems. At last the law has a social impact because of more advanced technology affects the society.  (https://www.intel.com/content/www/us/en/silicon-innovations/moores-law-technology.html) Another reason for this prosperity can be explained by looking at the IT companies and their business model. Google provides most of their services free of charge but the user generates valuable data when using the service. Google has compiled


The Search: How Google and Its Rivals Rewrote the Rules of Business and Transformed Our Culture



you might say that we went from an AI winter to AI spring. In the following article, the future of AI will be explored. The focus will be the humanization of AI and focuses on how people and the society will interact with AI.

Before going too much into detail it is necessary to understand the fundamentals first.

AI is an abbreviation of “artificial intelligence” which is basically just a term for anything where there is an automated decision being made. It could automatic doors at an entrance or a coffee machine brewing a cup of coffee with a push of a button.

When we are talking about artificial intelligence assistant professor of Michigan State Universty, Arend Hintze defines 4 types of AI:

Type 1 – Reactive Machines

This type of machines reacts to a state without memorizing it or learning from it. This is the most narrow kind of AI since it would do the same decision given a certain set of criteria and if these criteria are not present the decision won’t happen. The examples from before with the automatic doors and coffee machine are examples of this kind of AI. But the category also includes AI such as IBM’s DeepBlue which were the chess playing AI that defeated former world champion, Gary Kasparov. The DeepBlue Machine analysed multiple chess plays and by that, it made a reaction pattern. This means that every time a certain state occurred on the chess board it would do the same move independent from the opponent’s previous moves. So in that sense, it wasn’t very different from the automatic doors.

Type 2 – Limited Memory

This type of AI can track variable over time and store them in a temporary memory which makes them able to adapt to the world around. If DeepBlue had a temporary memory it would have been able to do the same as before but also adapt the move of the chess pieces according to a few previous.

Type 3 – Theory of Mind

These machines would have an internal concept of the world giving them a limited understanding of inhabitants of this world. In another way, they would awareness of others but this is not strong AI which will be explained in the following type of AI. If AI should even have a chance of cooperating with us, it has to have an understanding of our behaviour and how we interact with each other. Autonomous cars are located in this category because they have to be aware of the others and especially humans in its interpretation of the world, but they haven’t fully reached an understanding of humans and our behaviours. A good example of this is when a Google Car assumed that a human driver would give space which he didn’t.

Type 4 – Self-Awareness

Strong AI would be when the AI reaches the singularity which is when machines will self-improve with such speed that the present state of a machine is irrelevant. AI in this category is aware of the humans and their emotions but also self-aware. Even though machines with recursive self-improvement has been seen recently it is only a step in the direction of full self-awareness. This type a been depicted in multiple Hollywood movies and Tv-shows but is at the moment still far away.

Since type 4 and some of type 3 still is science fiction we will focus on type 1, 2 and a small bit of type 3.


Before we can talk about humanizing AI the article will quickly go through the termonology used when talking about developing AI.

All the types mentioned before are types of AI. ML or Machine Learning is when the AI learns how decisions are made in order to make predictions.

A prediction could be made by looking at a photo of a person in a photo album and from this the ML will learn. The ML will now predict which of the other pictures from the album has the same featured.

Another example is autonomous cars which observe the traffic and because of driving experiences it has, it can predict that when there is a red light ahead it should stop. Or at least in theory.


How does it work?:


Qualities/details of the data


Features of a movie could be the length of the video, the colours used for the video, size of the video, it was recorded, etc.

Features of a tiger could be the gender, number of stripes, the weight, etc.

Features of text are bit more complicated but essentially it boils down to combinations of words and from these combinations, the machine can learn how to use the words.

Features of faces could be, eye size, the length between nostrils, hair length etc.


A 0-n value. This value is an expression of how confident the machine is in it’s prediction.

When we talk about classification we have 4 different types:

The machine predicts whether or not the data belongs in a category.




Exciting Services and companies:

TensorFlow: Open Source AI made by Google. Has large sets of data available and libraries for easy coding.

Adobe Sensei: Adobes new AI helping in Adobe CC

Adobe XD: New tool for developing APPs just by drag and drop.

Teachable Machine: Website that allows the user to quickly develop an AI that can distinguish between three states using image recognition.

STØJ: Creative Coding Workshop


With the rise of AI many questions has been raised such as: Will AI take our jobs? How will AI shape the future of APP development? To answer all of these questions we need to understand the principles of AI and how it interacts with it’s environment.





Leave a comment

Your email address will not be published. Required fields are marked *