Jack Morris
Session (1)
Jack Morris recently completed his PhD at Cornell Tech studying the information theory of language models, with pioneering work on embedding inversion (vec2text), contextual document embeddings, and language model capacity/memorization — making him a compelling voice on the science of what language models actually know and how retrieval systems work.
Recent talks (1)
GitHub
@jxmorris12- jxmorris12/language_tool_python★ 524A free, non-AI Python grammar checker
- jxmorris12/cde★ 205Code for training and evaluating Contextual Document Embedding models
- jxmorris12/bm25_pt★ 106Minimal PyTorch implementation of BM25 with sparse tensors
- jxmorris12/gptzip★ 80Losslessly encode text with arithmetic coding and HuggingFace Transformers
- vec2text/vec2textorg★ 1,121Utilities for decoding deep representations (like sentence embeddings) back to text
Recent writing (5)
- Superintelligence from First Principles · blog · Jun 2025
- The Case for More Ambition · blog · Jun 2025
- 'AI' just means LLMs now | Superintelligence Part 1 · blog · Aug 2025
- Harnessing the Universal Geometry of Embeddings · paper · May 2025
- How much do language models memorize? · paper · May 2025
Podcasts & interviews (2)
- Information Theory for Language Models: Jack Morris · Latent Space: The AI Engineer Podcast · Jul 2025
- Jack Morris on Finding the Next Big AI Breakthrough · Odd Lots (Bloomberg) · Sep 2025
Public activity researched automatically · as of Jun 2026