If it works, it's no longer AI

This famous sentence comes from one of the forefathers of AI research,John McCarthy. It still holds true over 50 years later.

The hype surrounding AI in Germany is growing all the time. Many people are asking themselves the question: will there really be software with its own consciousness - artificial intelligence - in the near future? Or are the success stories just a manifestation of the current hype?

The clear answer is: neither!

No other software technology is currently as underestimated and overestimated as AI. Why?

We love magic! Although we know that real magic doesn't even exist, we attend the numerous shows by the millions. When the supposed magic behind the scenes is revealed to be simple mechanics, we are very disappointed. Deep down, we humans have a longing for real magic.

It is very similar with artificial intelligence: we have always and in all cultures been fascinated by the idea of intelligent machines, an artificial intelligence, whether good or evil. But when we realize that the seemingly magical effect of AI is basically just based on mathematics, our disappointment is great. We secretly dream of more.

What AI can already do today

The goals of the field of artificial intelligence, which was founded back in 1956, would probably be described today asmoonshot thinking: the idea of being able to create something completely new and revolutionary.

However, the successes achieved so far are quite remarkable and have already had a lasting impact on our everyday lives and software development. Be it chess computers, search engines, vehicle configurators, recommendation algorithms, spam filters, business rules or even navigation systems: Objectively speaking, all of these are (partial) results of AI research. Unlike at the time of their development, nobody would describe them as intelligent today.

All these technologies have long since arrived inthe computer science mainstream. In addition to the SIMULA-67 programming language,Marvin Minsky's thoughts on FRAMES from his 1974 articleA Framework for Representing Knowledge certainly form the basis for the later success of object orientation and many object-oriented programming languages.

The future of artificial intelligence

Theneural networks that are so topical and much discussed today essentially go back to a work byMcCulloch and Pitts from 1943(!). They already experienced a heyday in the 1990s and are now omnipresent in our homes, cars and factories - mostly already implemented in hardware - and have since been disguised as control technology.

The new neural and data-driven algorithms are now clearly superior to humans in terms of performance in some disciplines. Like all their predecessors, these processes are here to stay. They will change our software and perhaps even dominate its development.

But no matter what name we use to describe the variety of knowledge-based, data-driven and learning algorithms in our software in the future, true "artificial intelligence" in the sense of theoriginal definition from 1956 or theTuring Test from 1950 is still a long way off.

Artificial intelligence continues to be an abbreviation for "future computer science" in the spirit of John McCarthy and remains - as it was in the 1960s - an important driver and thought leader in software technology. As the past clearly shows us, this is "revolution" enough.