Artificial intelligence: Will we soon live with robotic humans who are smarter than us?

Hardly any other field of computer science triggers as many emotions as “artificial intelligence”. Will machines one day have a consciousness? Will they be able to think like humans?” These are the questions of our time. Many people feel reminded of intelligent humanoid robots, as we know them from science fiction novels or movies.

Today, clever machines (AI) as the strongest driver of digitalization and digital transformation.

Often we don’t know exactly what an artificial intelligence is, what it can and cannot do. Therefore their ability is regularly overestimated or underestimated. In the social Discourse, opinions move between exaggerated expectations, but excessive fears.

Mind and body

Who or what should think like a human being, also needs a body. After all, the human mind is also inseparably connected with the body. This is why robots play a decisive role in machine learning. AI’s that have a robot body can orient themselves and move in space. They should touch, lift or transport objects. The relationship between robot and AI – i.e. the cognitive thinking machines – is roughly comparable to that between body and mind.

Cognitive machines will completely change our world.

With their establishment, an environment is created in which we are increasingly will bypass smart machines. In the world of tomorrow, it will be normal to communicate with an AI, whether at work or at home. The artificial intelligence is gaining ground in all areas of our lives and will be become as normal as our refrigerator, TV or the Smartphone.

How do the machines get smart, how are they learning?

There are currently three strategies for machine learning: Supervised Learning, Unsupervised Learning and Reinforcement Learning.

Supervised Learning

Supervised learning is said when humans dictate and monitor the learning progress of the machine. Pre-categorized data is provided to the machine. Suppose the machine is supposed to master the detection of dog pictures. For this purpose, this data already categorized during supervised learning is presented – i.e. a lot of pictures of dogs. Based on the many examples, the AI begins to remember its own model of the optics of dogs in pictures. After learning, AI can be tested: how well can it detect the appearance of dogs in pictures?

Unsupervised Learning

In this case, so-called raw data are provided. So not, as in supervised learning, data pre-categorized by humans – like the pictures of the dogs. In unsupervised learning, the data is completely messed up, for example many colorful images of all kinds, without any assignment to a topic. The machine is fed with these raw image data and ideally finds categories and patterns itself.

In order to distinguish machine learning from human learning, it is important to understand that machines have no context or world knowledge. A machine does not know that dogs belong to the animals. She doesn’t know that the genus animals exist, even if it recognizes dogs in pictures.

When unsupervised learning goes well, AI finds visual similarities. For example, things like fur, light and dark and so on. In the best case, it filters out the category “Pictures of dogs” independently from the set of pictures.

The interpretations of the images are presented to humans and monitored during unsupervised learning. This is because AI’s do not distinguish between causalities (causes) and correlations (relationships), which carries the risk of misinterpretations.

Reinforcement Learning

So-called empowering learning is the last cry in machine learning and many expectations are associated with it. For certain applications, this learning strategy has proven to be very effective. In Reinforcement Learning, AI is stimulated with a kind of reward for continuous optimization. For example, games are well learned when the AI is given the goal of achieving the highest possible score.

For such an advanced learning system, therefore, the East Asian Game “Go” no longer a problem. At the beginning of the learning process, a random control order or move. If the AI is used to Point-scoring, she learns which train was successful and remembers This. The machine plays an incredible number of times before becoming a professional and, unlike humans, all traits and their success. That’s why Google AI beat the world’s best Go player.

You might think that the machine is smarter than humans, but that’s not true. This method of learning also has weaknesses. First of all, learning is limited to the respective application area – for example, the game rains of the Go game. Attempts to transfer the learned knowledge to other fields – for example with the knowledge of the GO – game also master the chess game, fall under the category “transfer learning” and are not yet particularly advanced.

Learning on the basis of a random train or trying out has so far also been only for relatively simple relationships, as in a game and its Rules.

If the relationships become more complex and the virtual environment offers more Possibilities and solutions, the AI sometimes finds no solution at all. It lacks world and contextual knowledge. Sometimes the learning progress also stagnates, because the system no longer recognizes any more rewards. Newer learning methods try to install a kind of curiosity on these AI’s, which they Continue learning motivated.

What do people and smart machines have together, what distinguishes them?

One thing is clear: we will inhabit the Earth together: a world without AI will no longer exist in the future – but it is not (yet) a competition for human intelligence. Complex deep learning networks now have dozens of layers of about a billion simulated neurons.

The human brain, on the other hand, has more than 86 billion neurons, each containing up to 10,000 other neurons (over synapses). When thinking and doing simple Things simply arise in our brains all the time without much effort new connections.

Even as the body and computers get closer and closer, there are areas of our minds that computers cannot simulate. These bastions of humanity include, for example, our feelings and our consciousness, the origin of which is still unclear. The question of whether artificial intelligence could also gain consciousness has led to a renaissance of research around human existence.

How it arises is an unsolved mystery, just as it is exactly what it is – a consciousness. Consciousness is presumed to be connected with the ability to self-reflection. In other words, the ability to see oneself through someone else’s eyes, to be aware of its effect on other people. This would mean that every conscious being has a mental representation of himself. However, more details are not yet known.

We understand what our consciousness does.

The attempts to copy and artificially create consciousness thus raise central questions of humanity itself.

First of all, we need to understand how intelligence works in humans in the first place. We know only one way of thinking – the human one. If we want to create an intelligence, we can only do so in its image. That’s why the way machines think can explain a lot about ourselves.

Experts from psychology, neurology and information technology are currently working together to explore human intelligence.

Concepts such as “thoughts” and “ways of thinking” are to be broken down and broken down into their smallest logical, chemical and electronic components. Many things of human thought are still completely in the dark. We don’t know how creativity is created, nor why our feelings are affected by certain substances that our brains release. We know the endorphins that our brains can release – but we don’t know why they make us so happy.

It will be a while before we get people serious competition. While we can’t remember what we’ve learned as reliably as the machines, we can do everything we can to relate to our environment and easily apply our knowledge and experience to other areas and creatively create completely new things.

Neither excessive expectation nor exaggerated fear is justified. Mindfulness when using smart machines, on the other hand, is certainly certain, because today we are setting the course for the world of tomorrow.

Smart machines can help us humans in many ways and take away annoying and monotonous tasks. However, you can abuse and how a weapon is directed against ourselves. In an ideal future, AI would take away from us annoying work and make everything resource-saving and produce in sufficient quantities. There would always be enough for everyone – neither too much too little.

Against the background of artificial intelligence, it is therefore always more important to ask what is actually the core of our human being and of our community.

Humanities and social sciences are experiencing a real flourish today: they have always been concerned with the origin of man, his nature and his values. How do we measure what is good and right for us? What use of artificial intelligence is justified when and – on the basis of what morality and what values?

These are all questions that we need to answer today in order to a world that is worth living in for all. How the future depends solely on our decisions in the present.

Kommentar verfassen