The Other Side of Complexity
Oliver Wendell Holmes tols us that while simplifications on this side of complexity are worthless; simplicity on the other side can be priceless. But getting there means entering the mess, learning what matters, then clearing the rest away.
Have we seen enough mess to clean up AI? Perhaps. Richard Sutton's "Bitter Lesson" offers the most complex path. Sutton says that scale wins — more compute, more data, always better. That works if you can afford it. Software engineering can't. Labels cost hours to days. An $800B gap yawns between AI revenue and datacenter bills by 2030, and the text that trained today's models runs out before the bill comes due.
So let us walk through the mess. Across 120+ real SE tasks, a minimal toolkit called EZR — 500 lines of Python, no deps, under 100 labels, under 10 parameters — performs as well or better than SMAC, TPE, and the heavyweight crew at a fraction of the cost. Not by being cleverer. By doing less, harder. Active learning picks informative rows. Naive Bayes sorts best from rest. A 20 line tree shows what matters. The other 990 variables out of 1000 get thrown away. They never mattered.
That's the other side of complexity. Earned, not assumed.
Speaker
Tim Menzies (NC State; prof, phd,)
ACM / IEEE / ASE Fellow
EIC of ASE journal
Bio
Prof Menzies earned his Ph.D. in Computer Science from UNSW (1995) and am is now a Full Professor at NC State University, where he directs the Irrational Research Lab (mad scientists 'r us). His work focuses on software engineering for AI, specifically building data-driven, explainable, and intelligent software systems. He is am an ACM, IEEE, and ASE Fellow, and co-creator of the PROMISE repository, showing that small, interpretable AI models can often outperform larger, more complex ones.
He has have published over 300 papers with 24,000+ citations (h-index=74) and advised 24 Ph.D. students. He serves as Editor-in-Chief of the Automated Software Engineering journal and, from 2010 to 2026, as an Associate Editor for IEEE TSE.
His research is supported by $19M in competitive grants from government (NASA, NSF, NSA) and industry (LexisNexis, Microsoft, Meta). Recent: $2.5M from NSF + $1M from NSA for trustworthy compact AI. For more details, see http://timm.fyi
Mon 6 JulDisplayed time zone: Eastern Time (US & Canada) change
09:00 - 10:30 | KeynoteKeynote / Research Papers at MB 2.210 Chair(s): Ali Ouni Ecole de Technologie Superieure (ETS) | ||
09:00 10mTalk | Opening Remarks Keynote | ||
09:10 60mKeynote | The Other Side of Complexity Keynote Tim Menzies North Carolina State University Pre-print | ||
10:10 20mResearch paper | Adaptive Software Energy Feedback for Long-Term User Engagement Research Papers Thomas Zaragoza Université de Pau et des Pays de l’Adour, Adel Noureddine University Paris Nanterre, Ernesto Exposito Universite de Pau et des Pays de l’Adour, E2S UPPA, LIUPPA | ||
10:30 - 11:00 | Coffee BreakFSE Catering | ||
10:30 30mCoffee break | Break FSE Catering | ||
11:00 - 12:30 | Research Session 1Research Papers / Hot-off-the-Press Track at MB 2.210 Chair(s): Wesley K.G. Assunção North Carolina State University | ||
11:00 20mResearch paper | A Simulation-Based Genetic Algorithm to Find Configuration Anomalies in UAV Software Research Papers | ||
11:20 20mResearch paper | Genetic Programming for Self-Adaptive Auto-Scaling of Microservices Research Papers Jia Li University of Ottawa, Mehrdad Sabetzadeh University of Ottawa, Shiva Nejati University of Ottawa Pre-print | ||
11:40 20mResearch paper | Prompt Optimization for LLM Code Generation via Reinforcement Learning Research Papers Ali Mohammadi Esfahani Carleton University, Nafiseh Kahani Carleton University, Samuel A. Ajila Department of Systems and Computer Engineering, Carleton University | ||
12:00 15mShort-paper | Search-Based Generation of Complex Inputs with FANDANGO Hot-off-the-Press Track José Antonio Zamudio Amaya CISPA Helmholtz Center for Information Security, Marius Smytzek CISPA Helmholtz Center for Information Security, Andreas Zeller CISPA Helmholtz Center for Information Security DOI Authorizer link File Attached | ||
12:30 - 14:00 | LunchFSE Catering | ||
12:30 90mLunch | Lunch FSE Catering | ||
12:30 - 14:00 | LunchFSE Catering | ||
12:30 90mLunch | Lunch FSE Catering | ||
14:00 - 15:30 | |||
14:00 10mTalk | Challenge Overview SSBSE Challenge | ||
14:10 15mShort-paper | ApkFuzz: Search-Based Fuzzing for Android APK Vulnerability Discovery SSBSE Challenge Karine Even-Mendoza King’s College London, Aidan Dakhama The University of Edinburgh, UK, Harel Berger Ariel University Pre-print | ||
14:25 15mShort-paper | Fuzz3: Entropy as a Third Oracle SSBSE Challenge Karine Even-Mendoza King’s College London, Janine Obiri University College London (UCL), Aidan Dakhama The University of Edinburgh, UK, Phil McMinn University of Sheffield, William Langdon University College London Pre-print | ||
14:40 15mShort-paper | Reducing LLMs by Searching Heterogeneous Quantizations SSBSE Challenge Enrique Alba University of Malaga, Yazhuo Cao King's College London, Hector Menendez King’s College London Pre-print | ||
14:55 15mShort-paper | SafeTune: Search-based Harmfulness Minimisation for Large Language Models SSBSE Challenge Giordano d'Aloisio University of L'Aquila, David Williams University College London, Giusy Annunziata University of Salerno, Zhiwei Fei Nanjing University, Antinisca Di Marco University of L'Aquila, Federica Sarro University College London Pre-print | ||
15:10 15mShort-paper | Survival of the Stealthiest: Evolving Low-Entropy Ransomware via Genetic Algorithms SSBSE Challenge Efrat Levenberg Jerusalem College of Technology, Kristina Sviazhina , Pierre Parrend Université de Strasbourg, Ayelet Butman Holon Institute of Technology, Harel Berger Ariel University | ||
15:30 - 16:00 | Coffee BreakFSE Catering | ||
15:30 30mCoffee break | Break FSE Catering | ||
Accepted Papers
| Title | |
|---|---|
| Opening Remarks Keynote | |
| The Other Side of Complexity Keynote Pre-print |