Why AI Does Not Eat Computer Science
2026/03/25 by Prof. Dr. Carsten Binnig, Data and AI Systems Lab
In recent months, a new narrative has gained traction: AI will “eat” computer science. The reasoning seems straightforward—if AI can generate code from natural language, why would we still need computer scientists? This view, however, overlooks a fundamental point. Computer science has always been about raising the level of abstraction to make programming more productive.
AI is making computer science more essential than ever
Computer science is not being replaced. On the contrary, we are witnessing a phase of acceleration. From early machine code to modern high-level languages, every step has made it easier to instruct computers. AI is simply the latest step in this long evolution. Yet, increased productivity has never eliminated the need for expertise – in fact, it has reinforced it.
Even if AI can generate code, someone must still determine whether that code is correct, efficient, secure, and reliable.
As software development becomes ever faster and more accessible, the need for sound technical expertise will only increase. Even though AI can generate code, someone still needs to decide whether that code is correct, efficient, secure and reliable. Without this understanding, AI-generated code remains a black box that users try to ‘persuade’ rather than truly control. The actual benefits of AI-generated code can only be reliably utilised by individuals with a solid grounding in computer science.
AI opens up new areas of application
Furthermore, AI is opening up many new areas of application. We can now develop specialised software solutions even in areas where the effort would not have been worthwhile in the past – for example, because the market was too small. It is now increasingly viable to create software even for one-off tasks, as implementation has become significantly faster and more efficient. Another advantage is that small and medium-sized enterprises (SMEs) – traditionally a strong sector in Germany – will in future be able to carry out projects with a small development team that previously only large corporations could manage.
At the same time, it is unlikely that programming using natural language alone will be the definitive paradigm. Many have already experienced how inefficient it can be to repeatedly refine command prompts in order to achieve the desired result. Natural language is inherently ambiguous, whereas complex systems require precision. The future of programming will therefore combine natural language with traditional, formal methods – in other words, what we know today as programming languages. This hybrid approach will require computer science knowledge rather than replacing it.
The future of programming will combine natural language with classical, formal methods. This hybrid approach will require, not reduce, strong computer science expertise.
AI also enables an important shift in how we think about software. Traditionally, software is static: it performs exactly the operations it was explicitly programmed to execute. AI-driven systems, in contrast, can become adaptive—supporting more flexible, human-like interactions and allowing variations that were never explicitly predefined. However, this shift challenges many established principles. For example, classical testing assumes fixed behavior, whereas adaptive systems may evolve over time. Ensuring correctness, reliability, and trust in such systems is therefore significantly more complex and requires new advances in computer science.
Computer science skills will thus be particularly relevant in a world shaped by AI. However, it is important that we further develop the skills of Computer Scientists and how a modern Computer Science should look like so that we can responsibly harness the full potential of this technology.
Prof. Dr. Carsten Binnig
Carsten Binnig is a LOEWE Distinguished Professor heading the Data and AI Systems group and is currently Dean of the Department of Computer Science. After completing his PhD at the University of Heidelberg, he initially conducted research as a postdoctoral researcher at ETH Zurich. He then spent several years in industry as a software architect at SAP before moving to the prestigious Brown University in the USA in 2014.
Among other roles, Binnig is Principal Investigator at the Hessian Centre for Artificial Intelligence (hessian.AI) and the Cluster of Excellence Reasonable AI at TU Darmstadt. He also heads the research area Systemic AI for Decision Support at the Darmstadt site of the German Research Centre for Artificial Intelligence (DFKI) and is a member of the ELLIS Unit Darmstadt. .
