How to make an AI robot

Machine learning dossier: the future is today

Lately the development of machine learning has accelerated: computers are getting smarter and smarter. But you don't necessarily have to be afraid of the talented machines.

"Venus and Adonis" by Rubens, seen from the algorithm. Information on the genesis at the end of the text.

Artificial intelligence (AI): That sounds like the future, like smart robots that chat with us about art over dinner. Or machines that serve us or, as some fear, turn against us like the Terminator. In our imagination, artificial intelligence often has a face, a shape, but in reality it is mostly shapeless; Software, algorithms that make decisions independently without needing our help. And: You have long been with us. In the economy in particular, the triumphant advance of clever technologies does not seem to be stopped any longer. The Japanese industrial robot manufacturer Fanuc, for example, has started equipping its machines with AI. The software comes from Preferred Networks in Tokyo. With them, Fanuc's metal gripper arms, usually in yellow, are able to learn, so they can, for example, teach themselves how to organize objects at lightning speed.




>>> Is there artificial intelligence with consciousness? >>>

What are the potentials and the limits of artificial intelligence? Susanne Biundo, professor at the Institute for Artificial Intelligence at Ulm University, knows the answer. In the enkelfähig interview, she explains why one shouldn't be afraid of the clever machines.




"The Adoration of the Magi" by Botticelli, interpreted by an algorithm. Information on the genesis at the end of the text.

Robots tell their “colleagues” what they have learned

This happens through sensors and cameras that record the process and thus collect data about it. Using this data, the robot learns its own tasks without having to program it beforehand. And that overnight, in eight hours, as Fanuc reports. And if a robot has a new skill in its repertoire, it can also teach its colleagues in the machine shop via a network. You pull the information from a data cloud. Software that Bosch Rexroth uses to better assess the wear and tear on machines in paper mills or steel mills works in a similar way. The software uses sensors on the machines to analyze the patterns of wear and tear and can therefore predict more precisely when parts and components need to be replaced.



>>> Will people soon be dispensable on the job market? >>>

Fast, smart, inexpensive: In the economy, machines could soon be doing more and more human tasks. But how exactly will that affect the world of work? Three theses from AI expert Philippe Lorenz.




"Judith with the head of Holofernes" by Guercino, seen by the algorithm. Information on the genesis at the end of the text.

Machines soon reached the level of humans

AI is also being used more and more in other areas, such as diagnosing diseases. A team of researchers from Beth Israel Deaconess Medical Center and Harvard University recently unveiled software that can identify breast cancer on diagnostic images. She's not quite as precise as a pathologist, but almost. It shouldn't take very long for the machine to reach human level. And yet it will not replace him. Rather, she becomes his assistant. Man gives the machine an order, explains what he is looking for, what his goal is, and it helps him to achieve it faster than before.




>>> The perfect pretzel - baked by the robot >>>

The German start-up PreciBake teaches ovens how to bake properly, thereby laying the foundation for intelligent kitchen appliances. You can now taste the first product: the perfect pretzel.




"Judith beheads Holofernes" by Caravaggio, interpreted by an algorithm. Information on the genesis at the end of the text.

The human feeds the machine with information, the software spits out hundreds of designs

The person becomes a “mentor”, as the technical director of Autodesk, Jeff Kowalski, puts it. Autodesk is one of the leading design software companies. They are already working on a program that is supposed to make creative decisions independently and that becomes more precise and more imaginative with each task. It's called "Dreamcatcher". The designer feeds the “Dreamcatcher” AI with information, for example about a bicycle that he would like to manufacture. The software spits out hundreds of possible designs within a few hours. The designer now selects what corresponds to his ideas. Ultimately, the creative's intuition remains decisive for the development of the product. It cannot be replaced. It's different with other jobs.




>>> Why artificial assistants are better than humans >>>

The New York company has developed an artificially intelligent personal assistant. Dennis Mortensen, CEO of, thinks they are significantly better than human team assistants. Is that the end of the classic secretary?



Classic works of art interpreted by algorithms

How do machines see art? To answer this question, the Italian scans Artist Quayola historical works of art, processed by a specially developed algorithm and then presented in a new form. In his “Captives” series, which we use to illustrate this dossier, Quayola shows a number of these 3D works of art. Here (from top to bottom, the original on the left and the machine interpretation on the right) the works “Venus and Adonis” by Rubens; "The Adoration of the Magi" by Botticelli, "Judith with the head of Holofernes" by Guercino and "Judith beheading Holofernes" by Caravaggio.