Software is the most developed means of production in our society.

Companies like Google, Facebook or Microsoft draw enormous power from the programming skills of their staff.

If you want to realistically assess this power beyond the usual hype, you should try to understand the principles on which Google's search, Facebook's selection of posts or the Siri and Alexa assistance systems are based.

The Dutch mathematician and philosopher Stefan Buijsman wants to help his readers with this, and he succeeds.

With “Ada and the Algorithms” he has presented a concise summary of essential technologies from the field of artificial intelligence that is understandable for laypeople and that does not neglect the social and political dimensions of this complex of topics.

It is not easy to decipher emotions

Buijsman demystifies the common technical terms.

It explains in simple terms what an algorithm is and what programmers and computer scientists understood by artificial intelligence at different points in IT history.

He also shows step by step how engineers try to recreate the human sense of sight with the help of neural networks.

Buijsman dedicates a separate section to the weak points of neural networks.

Basically, it provides the essential arguments for the existence of organizations like Algorithm Watch, which want to make automated decision-making systems as transparent as possible.

The question of who trains and uses the neural networks with what material and for what purposes has long since left the purely academic field.

Algorithms as lightning rods

Buijsman describes the field test of a lie detector called "Silent Talker", which was supposed to detect false statements by test subjects by analyzing their voices and facial expressions. "Silent Talker (...), Which was tested at European border crossings in the summer of 2019, was correct in 75 percent of the cases in a group of 32 people," he writes. 75 percent, that sounds like a success, but an extensive overview study found no evidence that you can tell a person's feelings by the expression on their face. A frown might or might not mean that someone is upset. "Actually deciphering emotions turns out to be extremely difficult: it is not enough to present a series of images or videos to a neural network and hope that they will be put in the right drawer."

For large organizations, in which the task is often to pass on the responsibility for delicate decisions like a hot potato, the use of such assistance systems, designed according to supposedly objective criteria, is of course more than seductive: "We would have liked to help them, but the computer says no, Unfortunately there is nothing we can do anymore. ”The algorithms then fulfill a similar lightning rod function as the deployment of management consultants before any savings measures planned anyway, only that the purposes and methods are still transparent.

Let the intelligent systems do the work

The play of forces in a neural network, on the other hand, is not always comprehensible, even for professionals. Buijsman emphasizes that neural networks make decisions in a non-transparent way: “Assuming an algorithm does something strange, how do you find out what's behind it? Who can be held liable if nobody knows exactly what a decision is based on? ”It is precisely this blurring, which on the surface appears as a kind of autonomy of the system, that makes it attractive to contemporaries who are referred to and paid as decision-makers. The temptation to use the “intelligent” systems as a kind of decision-making washing machine in order to shift responsibility onto them is great.

Another important lesson for Buijsmans is based on the old data worker slogan "Garbage in - Garbage out" (GIGO). If a neural network has been fed insufficient and / or incorrect data, it will not be able to present any helpful results. This in turn is the cause of the insatiable greed for data of some institutions: "It is precisely these network limitations that mean that companies and authorities want more and more data from us."

Despite all the well-founded skepticism, Buijsman is convinced that the technologies summarized under the term artificial intelligence are predominantly helpful in many professions and life situations.

The software will not replace people in the foreseeable future, but will support them in their work.

Buijsman's examples also make it clear that the decisive questions are ultimately less of a technical or mathematical nature, but rather concern the power structures of society.

Stefan Buijsman: "Ada and the algorithms".

True stories from the world of artificial intelligence.

Translated from the Dutch by Bärbel Jänicke.

CH Beck Verlag, Munich 2021. 236 pp., Ill., Hardcover, € 20.