Meteorologist, oceanographer, Python solvem probler, experimenter, Raspberry Pi tinkerer, geek dad, manager, space enthusiast, ham radio lover, science columnist, ADHDer, synthetic intellect curious, linux admin cosplayer.

Nathaniel Rochester III

Born: January 14, 1919 Died: June 8, 2001 (aged 82)

Education

  • B.S. in Electrical Engineering, MIT (1941)
  • Elected to Tau Beta Pi and Sigma Xi honor societies

Early Career: Radar

After graduating from MIT in 1941, Rochester spent three years at MIT’s Radiation Laboratory (Rad Lab), the wartime center for radar development. He then moved to Sylvania Electric Products, where he was responsible for the design and construction of radar sets and other military equipment.

His group at Sylvania also constructed the arithmetic element for MIT’s Whirlwind I computer – connecting Rochester to Jay Forrester’s pioneering project.

IBM Career (1948 onward)

Rochester joined IBM in Poughkeepsie, New York, in 1948. His contributions were foundational:

IBM 701 (1952)

Rochester co-designed the IBM 701 – IBM’s first mass-produced scientific computer – with Jerrier Haddad. The 701 (originally known as the “Defense Calculator”) was announced in April 1952 and became the machine that established IBM as a dominant force in commercial computing. Rochester was the architect of its arithmetic unit, for which he received an Outstanding Invention Award.

First Symbolic Assembler

Rochester wrote the first symbolic assembly program for the IBM 701, allowing programmers to write instructions in short, readable mnemonic commands rather than raw binary or octal numbers. This was a critical step in the evolution from machine language to higher-level programming – predating the later SAP (Symbolic Assembly Program).

IBM 700 Series

Rochester managed engineering for the entire IBM 700 series during the development of models 703, 704, 705, and 709, becoming chief architect of the series. He also received a patent for the tape processing machine’s variable-word-size architecture.

AI Research (1955–1960)

Pattern Recognition and Neural Networks

When IBM Research was established in 1955, Rochester joined and led a group studying pattern recognition, information theory, and switching circuit theory. The group simulated the behavior of abstract neural networks on IBM 704 computers to test Donald Hebb’s theories of neuronal learning.

In 1956, Rochester, John H. Holland, L. H. Haibt, and William L. Duda published “Tests on a Cell Assembly Theory of the Action of the Brain, Using a Large Digital Computer” in IRE Transactions on Information Theory – one of the earliest computational neuroscience papers. The simulations were among the first to test theories of how neural assemblies form and learn.

The Dartmouth Conference (1956)

Rochester was one of the four principal organizers of the 1956 Dartmouth Summer Research Project on Artificial Intelligence, alongside John McCarthy (Dartmouth), Marvin Minsky (Harvard), and Claude Shannon (Bell Labs). In 1955, the four co-authored “A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence” – the document widely credited with coining the term “artificial intelligence.”

Rochester co-secured a $7,500 grant from the Rockefeller Foundation to fund the workshop. During the summer conference, he commuted to IBM’s facilities in Poughkeepsie to run neural network simulations on the IBM 704, providing the workshop with computational resources.

The Dartmouth Conference is universally regarded as the founding event of artificial intelligence as a formal field.

AI Projects at IBM

Rochester supervised several pioneering AI projects at IBM:

  • Arthur Samuel’s checkers program: One of the earliest machine learning programs; it learned to play checkers at a level that could beat its creator
  • Herbert Gelernter’s geometry theorem prover: An early automated reasoning system
  • Alex Bernstein’s chess program: One of the first chess-playing programs

In 1958, Rochester visited MIT as a Visiting Scientist and assisted with the development of LISP, John McCarthy’s programming language that would become the lingua franca of AI research.

IBM’s broad support for AI research was curtailed around 1960 due to shareholder concerns and public fears about “electronic brains.”

Government Service

Rochester served on advisory panels addressing air defense, antisubmarine warfare, cryptanalysis, and air traffic control – leveraging his expertise in both computing and radar.

Later Career at IBM

  • 1960s: Directed research in cryogenics and tunnel diode circuits
  • 1961: Moved to IBM’s Data Systems Division; established a group that designed IBM’s first two timesharing systems (QWIKTRAN and CPS) and initiated PL/I language design
  • 1975: Worked on IBM Chord Keyboard development
  • Later years: Worked on portable personal computer development (noted in a 1981 IBM Journal biography)

Awards and Honors

  • IBM Fellow (1967) – the company’s highest technical distinction
  • Fellow, Institute of Electrical and Electronics Engineers (IEEE)
  • IEEE Computer Society Computer Pioneer Award (1984)
  • Featured in Time Magazine (May 11, 1981)
  • Outstanding Invention Award (IBM) for IBM 701 arithmetic unit

Personality

Rochester is less well-known than the other three Dartmouth organizers (McCarthy, Minsky, Shannon), but his contributions were arguably more practically grounded: he provided the hardware (IBM 701), the software (first assembler), and the computational resources (IBM 704 access) that made early AI research possible. He was the organizational and engineering backbone behind the theoretical visions of his more famous colleagues.

Connections to Others

  • Jay Forrester: Rochester’s group at Sylvania built the arithmetic element for Forrester’s Whirlwind I
  • John McCarthy: Co-organizer of Dartmouth Conference; Rochester assisted with LISP development
  • Marvin Minsky: Co-organizer of Dartmouth Conference
  • Claude Shannon: Co-organizer of Dartmouth Conference
  • Arthur Samuel: Supervised Samuel’s pioneering checkers program at IBM
  • Donald Hebb: Rochester’s neural network simulations tested Hebb’s learning theory
  • John H. Holland: Co-author on the 1956 neural network paper; Holland later became the father of genetic algorithms
  • Jerrier Haddad: Co-designer of the IBM 701
  • Gene Amdahl: Rochester managed the 700 series engineering; Amdahl was chief designer of the 704

Sources