AI Doomsayers

On a passenger liner traveling from New York to Buenos Aires, Dr. B. beats the reigning world chess champion in a game of chess. The game evokes memories of Dr. B.'s time in solitary confinement at the hands of the National Socialists and jeopardizes his health. This chess novella was written by Stefan Zweig (1881–1942) in exile in Brazil. It is his last and best-known work.

For many years, AI algorithms have learned from their mistakes through reinforcement learning, becoming masters of strategic games such as chess and Go. However, language models such as ChatGPT have not yet mastered this type of learning. Dr. B. employed human reinforcement learning during his time in solitary confinement under the National Socialists in Vienna to become a chess master.

Reinforcement learning provides feedback, prompting AI programs to question themselves and learn from their mistakes. Ultimately, the AI arrives at a correct and provable answer. This means that the dystopian predictions of AI doomsday preachers could come true. In an interview with the German news magazine Der Spiegel (29/2025), American researcher and former OpenAI employee Daniel Kokotajlo explained why AI could soon take over human activities and then turn against its inventors: "As soon as deception is no longer necessary, it will wipe out humanity."

The path has been laid out. The timing of this singularity is unknown, but, as with many things, it will follow exponential growth: first, there is one grain of rice on the first chess square; then, two; then, four; then, eight; and so on. Ultimately, on square 64, there will be so many grains of rice that Germany will be buried meters deep.

This could happen quickly, so the IT sector, including SAP, should start thinking about future business models. As Harvard professor Clayton M. Christensen (1952-2020) aptly put it in his book “The Innovator's Dilemma”, this disruption is not only painful but also difficult: "If a company tries to develop a disruptive technology to the point where it meets the needs of customers in established markets—which most leading companies do—it is almost certain to fail."

A reinforcement learning program could study accounting, controlling, and logistics, and additionally offer SAP customers a better ERP system, perhaps one that is open source and license-free. This disruption presents an innovator's dilemma and could be a doomsday scenario for SAP. Strictly speaking, it's only a matter of time before SAP shares become worthless, right?