Will humans ever feel sorry for robots?

Artificial intelligence

Thea Dorn

To person

is a writer and philosopher. Her novel "Die Unglückseligen" (Knaus) is about the abolition of mortality. The subject of artificial intelligence concerns her: At this year's annual meeting of the German Ethics Council on the subject of "Autonomous Systems", she examined the question of how these machines change our self-image.

Rául Rojas

To person

is a mathematician and professor of computer science at the Free University of Berlin. He heads the Dahlem Center for Machine Learning and Robotics and became known for robots playing soccer, with which he and his team won many championships. He is currently working on autonomous vehicles, the prototype is called "AutoNOMOS". His doctoral thesis at the time had a completely different topic: Karl Marx.

Rául Rojas and Thea Dorn on Artificial Intelligence - double interview

If you could wish for an intelligent machine, what would it be?

Thea Dorn - I want to work with my head and my heart, which is why I think it's wonderful that devices do household chores for me that I would only grind my teeth to do. But it has never occurred to me to call my washing machine intelligent.

Raúl Rojas - I want technology that doesn't constantly get in my way and that ties me to the computer - like emails do, for example. I don't have a cell phone, this is my way out. I want to use my brain for something else. So I'm developing, for example, autonomous vehicles that can be used like taxis.

Aren't we at the mercy of them?

Rojas - no On the contrary. I want to turn the car into a form of public transport. We have 1.3 million vehicles in Berlin, plus buses and subways. I would like to have only 100,000 autonomous vehicles instead of cars: shared taxis. They would pick up people on the street and take them to the subway, bus, or other destination. You drive together, share the costs. That relieves the traffic. There would no longer be any parking lanes, and the bicycles would be able to use the cleared lane. There would be fewer accidents. And it would be so convenient! I no longer need to drive myself, I can get into a taxi, read, work. That would be my "autopie" for the city.

mandrel - No objection to this vision! Ultimately, I don't care whether a chauffeur is behind the wheel or whether the car drives itself. But for the time being we're still behind the wheel ourselves - and we're constantly being patronized. It makes me mad when my car tries to tell me by hysterical beeping how to park or when to take a break. And navigation systems are the purest of orientation machines. If possible, I'll turn it all off.

Rojas - I don't have a navigation system either. But seriously: I give a lot of lectures about autonomous vehicles and the most common reaction is: I don't want someone to take the steering wheel away from me! I answer: Rich people have a chauffeur and don't care about the details. Don't we all want to be rich? Then I am no longer interested in driving. I read my newspaper, my book, until the chauffeur says: We have arrived.

mandrel - This too: beautiful vision! I just fear that people will not use the time gained to read Aristotle, or at least read the newspaper. They'll be fiddling with their devices even more. But it is of course true that - especially in big cities - traffic needs to be improved a lot. Everyone who is on foot or by bike in Berlin hates car traffic.


mandrel - I am concerned about the speed with which we delegate to technology decisions that we used to have to make ourselves or skills that we had to acquire ourselves: Developing a feeling for whether I have moved enough today? No need, the health app tells me. Inventory planning? Superfluous, the refrigerator knows itself when new milk is needed. Learn foreign languages? Why, there are translation programs.

Mr. Rojas, you want technology that does things for you and keeps work away from you. So let's imagine a household that washes laundry by itself, does the food ...

Rojas - No no. I don't like household robots. My personal sport is mowing the lawn, picking up leaves and doing the laundry. You shouldn't relieve people of everything in the household because they'll stop moving and get fat. Like in the movie "WALL-E". Do you know him?

mandrel - Just the trailer.

Rojas - People are sitting on a floating couch with the screen in front of their noses, and robots bring the food. Such a future would be cruel. Some machines that could be built should therefore not be built for social reasons. Care robots for hospitals, for example, that minimize contact between the sick and other people.

mandrel - Massive work is being done on their development. On the one hand, this is understandable: there is a shortage of nursing staff, so why not use robots to remedy the situation? On the other hand, with every technological solution to a social problem, our social creativity will continue to decline. In essence, it seems to me to be about: How do we fill in the freedom that the new technology gives us? Are we approaching an Athens 4.0, in which everyone can philosophize and do sports from morning to evening, because friendly robot slaves do all of the tedious work for them? Or do we end up as "WALL-E" suggests: fat, lazy and stupid?

Rojas - That is the dilemma. Classical antiquity is not my role model anyway, because only men were free, while women and slaves did the work. I think more of the utopian socialists.

mandrel - Trotsky would be a great advocate of technologization today because he would hope that it would give people the freedom to educate.

Rojas - The big question with the early socialists was who did the menial jobs. Your answer: machines. Friedrich Engels wrote when he was still young: You said we need a lower class that cleans the toilets, but now there is an automatic toilet that flushes itself! For him, this proved that humanity of the future would be released from these lowly jobs - the socialist utopia of the 19th century. We now have many more machines than Marx and Engels ever dreamed of, but we don't have a great future with free people. One can almost despair.

mandrel - Well I think people have only two options: Either they continue to resign themselves to the fact that in addition to all the great things they can do, they are also limited deficiencies: sometimes sad, sometimes sick, and in the end there is death. Or he wants to leave all of this behind for good and be forever fit, always in a good mood and, ultimately, immortal. Our occidental, humanistically shaped image of man would be at an end. I suspect that in Silicon Valley I would earn derisive laughter if I said there that I consider the further optimization of humans in machine style to be the end of our autonomy and dignity. The very boldest there dream of uploading their brains to the cloud, where it can merge with world knowledge: total dissolution of boundaries. In my language: the end of personality, character, individual.

Rojas - These are all fantasies, they are all nonsense.

mandrel - I am reassured when you say that.

As a normal reader, how do I know that this is nonsense?

Rojas - It's hard to find out. Ray Kurzweil, the mastermind of transhumanism, claimed decades ago that humans would be replaced by machines, that at some point there would only be machine intelligence. He says by 2030 computers will have as many transistors as humans have neurons, so machines will automatically be smarter. This is nonsense and it is presumptuous. Again and again we notice how far we are still from understanding the brain and cognition. Anyone who believes that machines can be exactly like people fails to recognize that we are body and mind.

mandrel - Kurzweil is so consistent that the body is an annoying appendage to the mind.

Rojas - If he could store all of his information on a hard drive, he would quickly find that he would not be the same without his body.

mandrel - As I said: I hope you are right and the "singularity" that Kurzweil dreams of will never exist. Still, I don't think that one should shrug off all trans- and posthumanist visions ...

Rojas - But. (laughs)

mandrel - ... because they confront us with the vanishing point of our technological history of civilization. They force us to ask ourselves: do we really want to end the path we have chosen so radically?

Rojas - Building robots is like a magician, and people are fooled. An example: if you let small robots push cubes around at random, a pile will form in the middle of the arena at some point. It's random, but people think: Ah! The robots build a tower together. You are intelligent!

mandrel - Isn't that because people have always dreamed of creating beings that they infinitely outperform - in health, longevity, beauty, strength, intelligence? Pygmalion and the ideal woman, Jewish mythology and the Golem, Nietzsche and the superman. So I'm not surprised by this enthusiasm. But I share your assessment that we are still dealing with rather cheap booth magic, at least at the moment. I was at a reception the other day and a little white robot was walking around, greeting the guests, and was able - mostly - to tell whether it was a man or a woman. If he thought he saw a lady, there was a rose.

Rojas - That is the stupefaction of the people. I know what the state of the art is. We all only cook with water. The people who build these robots pretend that the robot really understands a lot. Instead of explaining the limits and problems of technology, they exaggerate excessively in order to get investors or new money for research.

mandrel - Do you think that these Artificial Intelligences could one day develop something like emotions?

Rojas - No. That is absolutely unthinkable. I've been working on it for 35 years. There are certain humble skills that we try to teach computers. But: the speech recognition is still not great. There are a few useful things like facial recognition at the airport or translation systems. The latter work because there are now so many books online in German and English that the computer can automatically learn the correspondence between the two languages.

This is called deep learning - people train the programs with large amounts of data.

Rojas - Yes. It only works to a certain extent, however. I show the computer millions of pictures of cats. Then at some point he will know what a cat is. But when I ask him why? He can't explain it.

Where does this data come from?

Rojas - From all of us.

mandrel - Every day.

Rojas - If you upload pictures to Instagram, the data ends up on Facebook. This allows them to improve their face recognition programs. Or cat recognition programs. We generate this data - voluntarily and free of charge for Google and Facebook.

A Google AI recently called black people gorillas. Another AI connected the kitchen and women. Do the systems not learn in a value-neutral way?

Rojas - No. With deep learning there are only patterns. And that results in a decision. If a woman is featured in 90 percent of kitchen advertisements, then it is clear where that comes from. Big data contains all of our values, contradictions and prejudices.

mandrel - I heard that one of these so-called autonomous systems in the military field, which tanks recognized well, suddenly failed. Nobody knew why. Then it was found that the majority of the images used to train the system showed tanks under blue skies. Now the sky was overcast and the system at a loss.

Rojas - Unfortunately, this is a fairy tale, such a system has never been used. A serious example is a Tesla accident because a real, commercial system was involved. Then a white truck blocked the highway. The car's camera couldn't distinguish the clouds in the sky from the truck, and it didn't brake.

mandrel - Nothing can replace human judgment. Man is able to weigh up and come to a judgment, to argue and to justify his decision. Can that be one of the differences between Artificial Intelligences and us?

Rojas - Yes. In fact, machines do not have deep judgment. Above all, they have no intuition, no tact.

mandrel - There is much more to it than that: Artificial intelligences know no feelings of shame or guilt, no pity, no conscience. The question that is often asked of your research area is: Who is responsible if a self-driving car decides in a precarious situation to run over the retired couple rather than the mother and child? Shouldn't we actually ask what it means when hundreds of thousands of vehicles are soon on our roads who simply don't care if they run over someone?

Rojas - I keep hearing this question: Who is responsible? The people who build the system, of course! My suggestion is to work so well in advance that such situations cannot arise in the first place. Minimum distance, reduced speed, lightweight construction materials - if we take all of this into account, we can hopefully build vehicle systems that do not cause serious collisions and are ultimately safer than today's cars.

Should AI be regulated internationally?

mandrel - Mankind founded the International Atomic Energy Agency when it realized what a dangerous and potentially earth-threatening power it had unleashed with nuclear power. It is probably time to think about whether we don't need a comparable facility to watch over the development and use of AI.

Rojas - I am always in favor of regulation, for example of genetic engineering and weapons. Only in this special case do I not immediately see how intelligent algorithms can be regulated. You have to think about it carefully - the solution is not exactly a trivial one.

And we really don't have to worry, Mr. Rojas, that robots will one day be smarter than humans or one day emancipate themselves from us?

Rojas - Definitely not in this century! In general, I don't think robots will take control of us at some point. Konrad Zuse, the first computer manufacturer in Germany, once said: If the computers get too cheeky, pull the plug.

The interview first appeared in "Chrismon".
Moderation: Mareike Fallet and Michael Güthlein, chrismon 12/2017, www.chrismon.de