Artificial Intelligence

Artificial intelligence (AI) is older than is commonly thought. The Turing machine, a precursor to the computer, was developed as early as 1936. The term "artificial intelligence" was first used at a conference of scientists in 1956. And the first chatbot saw the light of day as "ELIZA" back in 1966.

However, AI only began to establish itself in everyday life from around 2011, for example through Apple's voice recognition system "Siri". In addition to the ever-improving hardware, this was primarily due to the emergence of the machine learning approach. Instead of predefining formal rules for a computer system, here the system learns on its own based on available data.

Machine Learning and its future

The most important achievement of machine learning is pattern recognition: For this purpose, thousands of images, text elements or sound bites, for example, are fed into the software together with the correct interpretation. From this, the software "learns" the correct mapping and can then apply these patterns to future data sets.

The most advanced variant of Machine Learning is Deep Learning, which is based on so-called neural networks. At the software level, they replicate the structure of the human brain, or more precisely, the interconnectedness of neurons in the nervous system.

In this way, it should be possible for AI to analyze data sets that were previously difficult to capture. This applies in particular to sensory data, for example with regard to the interpretation of human facial expressions or body postures.

The benefits of artificial intelligence

AI is now used in almost all technological areas. Particularly important fields are:

  • Medicine: AI helps to make better diagnoses, for example by interpreting X-ray images more accurately than a human could.

  • Image recognition: AI can be used to identify, classify and optimize objects in images. A well-known example of this is Google Lens.

  • Speech recognition: Spoken language is converted into text using AI. AI is also used in translation programs such as DeepL. Text in images is also "read" by AI, making PDF documents searchable, for example.

  • Video technology: AI can automatically classify videos and recognize objects or even people in them. For example, criminals can be identified more quickly on surveillance videos.

  • Industry: AI can be used to automate many manufacturing processes as well as internal logistics. AI-based test algorithms reduce the error rate in manufacturing and thus increase the output ratio.

  • Financial institutions: AI has found its way into both the investment and insurance sectors, where it performs numerous analytical functions.

  • Automotive: As described above, AI is the backbone of autonomous driving.

AI: What is it all about?

These days, artificial intelligence (AI/AI) is on everyone's lips. Some conjure up the end of (human) labor, others are afraid that Skynet will soon become the new reality, and some are convinced that AI will solve almost all of humanity's problems in the near future.

Right off the bat, it's safe to say that all of these scenarios are highly unlikely. There are few technologies that have been hyped as much as artificial intelligence and, accordingly, the images in our heads are in many cases very far from what the current reality looks like.

Nonetheless, there are also few technologies that will impact all of our lives as fundamentally as AI systems are already doing. The potential, both good and bad, is insanely large and development is proceeding at a breakneck pace. It just simply can't keep up with the images painted in people's minds, movies and books.

AI: What is it all about?

It is completely impossible to cover the technological and social complexity of this technology in a short article. But at a very basic level, we should have at least a rough understanding of three key terms: Artificial Intelligence, Machine Learning, and Deep Learning.

The fabled artificial intelligence describes all systems (of a machine nature) that exhibit some form of "intelligence." Such systems, much like humans, can solve problems that require some learning and understanding. Very roughly, a distinction is made here between strong and weak artificial intelligence. While strong AI has general intelligence and can solve virtually "all" problems, weak AI is limited to individual problems and application domains, for example speech recognition.

In reality, there are a lot of weak AIs at the moment, but no (known) strong AIs.

A sub-discipline of artificial intelligence is machine learning. Machine learning encompasses a variety of methods and algorithms that allow computers to learn information from data and make decisions based on that information.

Two of the most common tasks in this area are classifications (Spam or no spam?) and predictions (How will the price change in the future?).

The methods used here are basically relatively easy to understand. In most cases they are mathematically well comprehensible algorithms.

A distinction is also made between supervised and unsupervised learning. In the first case, the computer is given "learning material", for example a series of examples of what a spam message might look like. In the unsupervised case, the system is supposed to draw information on the data without human intervention or material (training data).

Deep learning is now another sub-discipline of machine learning. Here, however, instead of relatively traditional algorithms, an attempt is made to model the functioning of human brains.

To do this, (deep) networks consisting of many layers which in turn consist of artificial neurons are simulated. Data then flows through these networks, much as it does in the brain. In the meantime, there are insanely complex network architectures that are capable of independently generating completely new insights from the data.

What does that bring us?

Various methods from the field of artificial intelligence are already being used in almost all areas. In principle, we can distinguish between three cases:

AI replaces a human and/or takes work from him or her. An example would be a self-driving car or a robot that relieves a human.

The AI does something that a human could basically do, but much faster and on a larger scale. This becomes exciting whenever the data volumes are very large. Humans can distinguish spam from important emails, of course, but there aren't enough humans to handle the large volume of data in a meaningful way.

AI does something that a human or humanity cannot. We are already seeing cases, for example in medicine, where AIs have found patterns in dates that were hidden from scientists.

In short, in the future we will rely on AI whenever the task is either too "boring" or too "big" (Big Data). However, it is of course also very likely that increasingly better AI systems will solve problems in the future that we are not even thinking about today.

On the other hand, there are many critics who are concerned that AIs could become a real threat to humanity. Here are four of the most important concerns and issues we need to keep in mind:

What happens if an "evil" human is the only or first to have access to an extremely powerful AI?

  • What happens when AI systems that may also have decision-making power learn the "wrong" things that are deemed morally unacceptable?

  • What happens when AI systems take social division to an extreme because those who have access to AI have an unlikely advantage over those who are "just" human?

  • What happens when we as humans can no longer understand what decisions an AI has made for us and why?

To get a handle on these issues we need to be proactive at various points. We need to make sure we all have a basic understanding of these technologies. We need to make sure that research and development is done publicly and that the technology is democratized. We need to make sure that we develop moral-ethical and legal standards to guide us.

Where do we stand?

As described above, we are still a long way from a general, powerful artificial intelligence that equals humans. However, we already have extremely powerful AI systems that make many fundamental decisions for us.

Whether we're getting recommendations on Netflix or Spotify, scrolling through Instagram, or somehow seeing advertisements, there's almost always an AI at play, learning from our behavior what we want or should watch next.

Many basic technologies, for example from the field of machine learning, have been so well researched and processed in the meantime that their use is relatively easy and inexpensive.

One exciting example is IBM's Watson, an AI that, among many other tasks, helps doctors in over 230 hospitals determine which cancer therapy is best for which patients.

Another example comes from the financial industry. Almost all major banks rely on AI systems to monitor transactions and keep track of the flood of bank trades. On the other side sit traders who use AIs to try to make profits in the stock market.

The list of examples could go on and on. The only thing that is clear is that AI is already being used in almost all products, at one point or another. If we don't play dumb, this shouldn't be a problem in the future, but rather a blessing!