MIT and AI: Research Labs, Project MAC, CSAIL, ELIZA

MIT's AI history includes the 1959 Minsky–McCarthy group, Project MAC, Weizenbaum's ELIZA and the 2003 merger that formed CSAIL.

kindUniversity
foundedApr 10, 1861

Updated

Research groups before CSAIL

The Massachusetts Institute of Technology was chartered on April 10, 1861. Its organized AI research began in 1959 when John McCarthy and Marvin Minsky founded an Artificial Intelligence Group within the Research Laboratory of Electronics and the Computation Center. Project MAC followed in 1963 with DARPA support for research in AI, computational theory and operating systems. The AI group therefore preceded Project MAC and the later CSAIL organization.

Joseph Weizenbaum's ELIZA paper appeared in the January 1966 issue of Communications of the ACM. Its first page identifies his MIT affiliation and the MAC time-sharing system. ELIZA analyzed input with keyword-triggered decomposition rules and generated replies through reassembly rules; a script presented in the paper imitated a Rogerian psychotherapist. This was a rule-based conversational program, not a modern learned language model.

The laboratories reunite

MIT's Artificial Intelligence Laboratory and Laboratory for Computer Science both grew from Project MAC and merged into the Computer Science and Artificial Intelligence Laboratory in 2003. MIT's current catalog describes CSAIL research across AI, computer systems and theoretical computer science, alongside teaching and collaboration. CSAIL is a research laboratory within MIT; its name is not projected back onto Weizenbaum's 1966 affiliation.

Sources

  1. [1]
    Charter

    The MIT Corporation · Apr 10, 1861

  2. [2]
    History

    MIT School of Engineering · retrieved Oct 24, 2026

  3. [3]
  4. [4]
    LCS and AI Labs Merge

    MIT CSAIL · Jun 30, 2003

  5. [5]
  6. [6]
    Computer Science and Artificial Intelligence Laboratory

    MIT Course Catalog · retrieved Oct 24, 2026

MIT history

2026

│
MIT researchers report GPT-6 Astra editing a small model's weights directly

Rikiya Takehi, Jayden Teoh, Alex Zhang and Omar Khattab of MIT CSAIL report that GPT-6 Astra, reading and editing weights without gradient descent, changed 0.016 percent of Qwen3.5-0.8B and raised its MATH-500 score from 7.9 to 48.1 percent. The authors call the results early and say the edit removed an obstacle rather than adding ability.

MIT CSAIL
Research

2025

│
MIT and Recursion release Boltz-2 code and weights

With Recursion

Researchers at MIT and Recursion released Boltz-2, which jointly predicts biomolecular structures and binding affinity. Code and weights were available under the MIT license; reported prediction benchmarks did not themselves establish clinical benefit.

Boltz team
Model

1966