ARTICLE

Johns Hopkins University

Created 2 May 2025
institutionuniversityresearchscaling-lawsphysics

Johns Hopkins University

Johns Hopkins isn’t the first name most people think of when they think “AI” — that’s Stanford, MIT, CMU. But it’s where Jared Kaplan did the scaling laws work that gave the entire industry its playbook. Sometimes the most consequential research comes from unexpected places.

AI-Relevant Work

Scaling Laws (Kaplan et al., 2020)

The most directly impactful contribution. Jared Kaplan, a theoretical physicist in the Department of Physics and Astronomy, brought a physicist’s mindset to AI: measure the system, find the governing laws, predict what happens at larger scales.

The result: proof that neural network performance follows clean power laws with compute, data, and parameters. This gave labs the confidence to spend billions scaling up. It’s one of the most cited and strategically influential AI papers of the decade.

Center for Language and Speech Processing (CLSP)

JHU’s CLSP has been a quiet powerhouse in natural language processing for decades:

  • Speech recognition research (pre-deep learning era)
  • Machine translation
  • Information extraction
  • Many NLP researchers trained here

Applied Physics Laboratory (APL)

The university-affiliated lab (one of the largest in the US) works on defence and security applications of AI — autonomous systems, signal processing, cybersecurity.

People Connected to JHU

  • Jared Kaplan — Scaling laws, Anthropic co-founder. Physics & Astronomy.
  • Jason Eisner — NLP researcher, CLSP
  • Mark Dredze — NLP, health informatics
  • Daniel Povey — Speech recognition (Kaldi toolkit creator)

Why It Matters Here

JHU represents how AI draws from unexpected fields. The scaling laws — arguably the most strategically important AI finding of the 2020s — came from a physics department, not a computer science one. The cross-disciplinary pattern keeps repeating: physicists, neuroscientists, mathematicians, linguists — all contributing to AI from angles CS alone wouldn’t find.

Go Deeper

enes