Artificial intelligence enriching the bright side of life
Trendwatchers are on it. According to Gartner Artificial Intelligence, Machine Learning, and Deep Learning will be the trends for 2017. A hot topic for meetups, blogs (Deep Learning, AI and the blackbox) and several community based movements and initiatives. Artificial Intelligence is intelligence exhibited by machines. Their interaction should mimic human behavior as much as possible, therefore they are designed to make the best possible decisions and draw conclusions based on their data analysis — that should make them far more efficient than humans.
Some believe that AI will be the end of the human race. Yet AI is coming and we are at the dawn of it. So, let’s take off our tinfoil hats and talk about the important things in life. Art. Music. And beer.
Art and Algorithm
The boundaries between art and technology are becoming more and more blurry. But what happens when we combine a mastermind’s creativity with Artificial Intelligence?
Thanks to Snapchat filters — which detect facial expressions and interact with it, using AI technology — we can share our daily creations with the entire world with two clicks. Among the many features, the AI interaction makes it possible to puke a rainbow when you open your mouth or turns your eyes into a destructive laser beaming weapon.
The other examples are photo editing apps. There are several photo apps that use AI algorithms for object recognition to recreate the target image in a certain artistic style, including well-known artists like Vincent van Gogh and Pablo Picasso.
While we were making selfies and calling ourselves true artists, Bas Korsten (executive creative director)and Emmanuel Flores (head of technology), at J. Walter Thomson had a bigger plan: The Next Rembrandt
This masterpiece –which is not an actual Rembrandt’s work, nor a copy of it– is the result. It also does not necessarily mean that Rembrandt would have painted it if he had still been among the living. So, how has this painting come to existence?
To answer this question, we need to know how they taught a computer to create a painting like Rembrandt? It has been done in four major steps: gathering the data, determining the subject, generating the features, and execution.
As Joris Dik, professor at TU Delft, described “there’s a lot of Rembrandt data available, you have this enormous amount of technical data from all these paintings from various collections”.
A facial recognition algorithm, identified and classified most of the patterns used by Rembrandt for painting human features. It learned its principles and replicated his style to generate a new painting.
An extensive database filled with Rembrandt’s paintings was built and examined, pixel by pixel. 150 gigabytes of digitally rendered graphics took more than 500 hours to complete and was gathered by a broad range of materials like high resolution 3D scans and digital files were analyzed and upscaled by deep learning algorithms to maximize resolution and quality. They noticed that most of Rembrandt’s art were created between 1632 and 1642. To further determine what the subject would be, demographics segmentation was analyzed. According to the results, 49% was female, 51% male. Further data lead to a portrait of a Caucasian male with facial hair, between the ages of 30 and 40, wearing black clothes with a white collar and a hat, facing to the right.
A software system was designed with the advice and assistance of Microsoft, Delft University of Technology, the Mauritshuis, and Museum Het Rembrandthuis to master his style. It was used to analyze and understand Rembrandt’s ways of using composition, geometry, his usage of pigments, and choice of material. A facial recognition algorithm identified and classified the most typical geometric patterns used by Rembrandt. Rembrandt is known as the master of light and shadow because of his innovative use of lighting to shape and paint human features. The algorithm even mastered the knowledge of deciding to either make these features sharp and in focus or soft and blurry.
To create this piece two big corporates (ING and Microsoft), a university, and a cultural heritage foundation came together –while they shared a passion for art, visualizing data, and AI– they used 3D technology to read the height and structure of each layer to mimic brush strokes, and scanned 168.263 paintings; therefore, I do call it ART.
Music-making Machines
The technique called Algorithmic composition offers possibilities to create music by using algorithms. Similar to the creation of The Next Rembrandt, it starts with creating a database –instead of paintings as input, it gathers data from songs. There are several models for algorithmic composition and categorizing the components of each song by their structure. The way of processing data is one of those. Examples of systems employing different AI methods based on their most prominent features, within this method of categorizing AI composition models, are the following subsections.
The research and development team of Sony’s Computer Science Laboratory in Paris choose the Hybrid system and their Flow Machines to create music with AI in collaboration with musicians by using algorithms. 13.000 leadsheets were analyzed to create one particular song. Since this article has a focus on combining technology with the finer things in life, this paragraph is dedicated to the most influential and popular act of the rock era, The Beatles.
FlowMachines is a system, comparable to the previous technology part on Art(??am I allowed to claim so), but instead of analyzing thousands of paintings, [you compared the music-making machines with the Next Rembrandt in the introduction] this system gathers data from a huge database of songs, from which it learns music styles, exploiting unique combinations of style transfer, optimization and interaction techniques –the system composes novel songs in many styles [you used the same phrases to describe Benoit Carré case ]. This makes it able to generate music autonomously or in collaboration with human artists. [this is a case — state it ] Benoit Carré –French composer and artistic director was responsible for the lyrics and producing the song, after it was composed by Artificial Intelligence. [link this sentence to the previous one]After “exploiting unique combinations of style transfer, optimization and interaction techniques, it is stated to be able to compose in any style.
FlowMachines is a system with a huge database of songs, from which it imitates, explores, and combines the style of others. It allows you to experiment with various constrains, until you forge your style. This makes it able to generate music autonomously or in collaboration with human artists.
Will music now become a computational object? The Music Industry is having a hard time, mostly because of technical innovation, [should be technological] what started with downloads and streaming and musicworkers (is it the right word?) are already trying to find new ways to sell their music. Artificial intelligence might be seen as a threat for the industry, Forbes even states that it might destroy it. In my opinion, even though musical style is the DNA from (of?)talented individuals, this development is still focused on keeping human as the central part of the creation, but FlowMachines AI provides it the possibility to further enhance music productions, originality and artistic creativity, mainly because the style of the musician using the system has the lead.
What now? Will music become a computational object?
The Music Industry is having a hard time, mostly because of the technological advancement. It started with the downloads and streaming and continues to this day as companies explore new ways to sell music. Artificial Intelligence might be a threat for the industry, even Forbes considered it “AI might destroy music”.
However, in my opinion, despite all the developments, humans remain the core of the creation. Even though FlowMachines AI provides the musical style –which is the DNA of the talented individuals– it gives the possibility to enhance music productions, originality, and artistic creativity. The musician’s style still has the lead.
Beer Made by algorithms
IntelliigentX, a startup in London, decided to add its own twist to the complex art of brewing beer. They have invented an algorithm that determines the amount of hops, water, yeast and grain in the recipe of the beer. Based on the preferences of the consumer, feedback on each version that comes from the users is adjusted by the algorithm. The bottles are labelled with links to the websites, , where you can provide feedback on your beer. The data is then used as input by the algorithm and to brew a better next batch.
The bots are comparable with the Facebook Messenger of a bot and asks the customer a series of questions concerning their opinion and experience with the by artificial intelligence produced beer. These questions can be answered with a rating between 1 and 10, yes or no or multiple choice possibilities.
The algorithm produces new recipes every month incorporating the feedback -levels of carbonation, bitterness and alcohol content all change based on how people are responding.
Because the system is constantly reacting to user feedback, beer that suits your taste can be brewed more quickly than any other method can. That means more data and a better beer to your taste.
Are all examples here AI? Is there a line between automation, assistants, bots and AI? What do you think?
