TedAI Vienna 2025 | Shaden Alshammari, MIT
Every day: 100+ new ML papers. New methods. New architectures. New „breakthroughs.“
But here’s the uncomfortable truth: We don’t actually understand how they fit together.
Shaden Alshammari from MIT just proposed something radical: What if Machine Learning needs its own Periodic Table?
The analogy is powerful:
In 1869, Mendeleev didn’t just organize elements – he revealed what was missing. His table predicted Gallium before anyone discovered it.
Now imagine that for algorithms.
The insight: Most ML methods aren’t truly novel. They’re variations of fundamental principles – like elements built from protons and neutrons, algorithms emerge from basic distributions.
Examples that clicked:
Clustering → guests gathering at tables
Dimensionality Reduction → forming a queue
What this unlocks:
✓ Hidden connections between „unrelated“ methods
✓ Gaps pointing to undiscovered algorithms
✓ A roadmap through the chaos
The question isn’t just „can we build this?“ – it’s „what could we discover once we do?“
What patterns have you noticed in the ML landscape that no one’s talking about?
#MachineLearning #ArtificialIntelligence #TedAI #Vienna #MIT #Innovation #DataScience #MLResearch
© TEDAI Vienna / Robert Leslie
Schreibe einen Kommentar