Super happy and very proud: JetBrains and TU Delft are launching an exciting five year collaboration: the AI for Software Engineering Lab (AI4SE).

There are five research tracks in this program:
- Validating (AI) generated cocde
- Optimizing code language models
- IDE-AI alignment
- Run time information in the IDE
- Programming education.

There are five PhD positions at TU Delft in this lab: Apply now!

https://lp.jetbrains.com/research/ai-for-se/

#jetbrains #tudelft #ai4se #icai @seresearchers

Today's #ASE2023 keynote on "resolving code review comments with ML".

Presented by Danny Tarlow (Google Deepmind).

Blog: https://blog.research.google/2023/05/resolving-code-review-comments-with-ml.html

Based on the DIDACT "Large sequence models for software development activities". https://blog.research.google/2023/05/large-sequence-models-for-software.html

#ai4se #codereview #ml4se #DannyTarlow #DIDACT

Currently working on a survey in the area of #AI4SE, I am seriously considering a "hall of shame" of the worst papers we found. It's truly unbelievable what kind of rubbish is published at some venues. The recipe seems: Take some random input representation, apply some random embedding, use some randomly chosen classifier, if you feel fancy use an attention mechanism (but don't say which one, let alone why), pick some subset of some badly biased data set, report success! Nothing can be learned from such work! Really, nothing. It's so disappointing.