Alexander Pretschner: Why GenAI Won't Replace Software Engineers
DIGHUM Lecture with Alexander Pretschner.

September 9th 2025
- 17:00 – 18:00 CEST
-
This is an online-only event.
See description for details.
On This Page
- Speaker: Alexander Pretschner, TUM & bidt, Germany
- Moderator: Carlo Ghezzi, Politecnico di Milano, Italy
About the Event
September 9, 2025
5:00 – 6:00 PM
(17:00) CEST
We are looking forward to seeing you:
Participate via Zoom (Password: 0dzqxqiy).
The talk will also be live-streamed and recorded on the DIGHUM YouTube Channel.
Abstract
Generating code with LLMs is surprisingly and increasingly powerful, prompting the question if we can expect humans to be out of the loop anytime soon. My answer is no to replacement for professional software engineers (and trivially yes to assistance). The argument is twofold and relies on two necessary distinctions: that software engineering is not just programming; and that software engineering ranges from writing scripts over implementing websites to building complex interconnected cyber-physical systems. The main argument is that software is a design artifact that embodies many architectural and technical decisions – traditionally taken (hopefully explicitly!) by engineers who understand and design the trade-offs. That is, some human needs to say what they want and which option they prefer – and this is not the realm of machines. Secondly, and once again depending on the context, software development projects often fail because it is outright impossible to state the requirements upfront. Agile software development processes embody that observation into an incremental process where what one wants is fully understood only while the system is built. And again, humans want the system to be built, not machines. A discussion of the influence of risk classes and productivity gains rounds off the argument.
Slides
Slides will be available for download after the lecture.
Video
Recording will be available after the lecture.