My view on the future

Based on my observations of the machine learning landscape today, with a particular focus on software engineering, I can see a few trends.

Language models will get better at completing software engineering tasks, such as requirements, testing, and documenting. This means that software engineers will be able to focus on their core work – engineering software – rather than on tedious, repetitive tasks. We will see models that will test software, document it, explain it, and maybe even repair it. The latest advancements in this field are very promising.

Hybrid models will be more popular. Combining symbolic analysis and neural networks will gain traction and be able to assist us in finding advanced vulnerabilities in software, as well as identifying them before they are exploited. This will make our software more robust and more resilient over time.

Large models and the availability of significant computational power will help us also to detect...