By: Larry Walker, August 2024; Context: AI, Knowledge Systems, ML, … .
From 1984 to 1987 Larry Walker led Sperry Univac’s initiative to bring AI Expert Systems technology to the commercial world. Two years after exploring the potential for Expert Systems, the Univac initiative was successful and moving fast. Key measures:
- A Five Year, $250 million initiative was endorsed by Sperry’s Chairperson, Gerry Probst and was making rapid progress.
- Univac OEM’ed AI hardware from Texas Instruments which captured MIT’s 30 years of AI research in their Lisp language-based computer.
- Univac purchased the marketing rights to Intellicorp’s Lisp-based operating system, KEE, which captured Stanford’s 30 years of AI research.
- Dr. Ed Feigenbaum, who created the Expert System technology at Stanford became a member of Sperry Univac’s Board.
- Over 350 Univac programmers were working on over 100 Expert System projects spanning every portion of Univac’s global enterprise.
- One of Sperry’s most senior programmers, John Marsden, reported that the KEE Operating System running on the TI hardware made him 100 times more productive than ever before.
- Fifty Univac customers had purchased this new AI environment and were working with Univac on custom solutions.
- Forty-three universities had received this AI environment from Univac and were applying it to training Univac’s future employees.
Larry remains certain that if Burroughs had not purchased Univac and stopped this 5-Year Plan, that Univac would have led the world into the dramatic solutions this technology enabled.
Larry and four of his Expert Systems team left the new Unisys and founded PEAKSolutions (PEAKS) which proceeded to deliver 39 of 40 contracted expert systems over a 5-year period. PEAKS’ success rested on its unique incremental development approach which it called Reflective Knowledge Engineering and preceded the Agile approach by years.
Route Builder (RB), custom developed for the State of Minnesota, was one of the most robust of these solutions. The State motivation for streamlining this process was to eliminate over 40 crashes into bridges per year. RB worked with the Transportation Department’s to map a safe route through Minnesota’s highway system for oversize, overweight vehicles. RB integrated hundreds of miles of highway detail, weight limits, height clearances for every bridge, and over 400 State laws which mandated the use of certain highways at certain times of the year. RB was the first solution for this complex challenge in the nation and was subsequently sold to 4 more States after Minnesota.
Despite its success, PEAKS succumbed to the AI Winter which struck in the early 90’s — primarily because many of the companies that rushed into the Expert System world failed to deliver on their promises. They had failed to discover the value of incremental development.
Larry then founded Knowledge Management, Inc. which developed a robust Knowledge Management Engine (KME) working with Nigel Dolby who had excelled at Expert System solutions with Univac. KME used a knowledge base to solve challenging problems. This approach moved programming to the knowledge level, eliminating the most common cause of software errors, the language disconnect between experts and programmers. KME got derailed by the arrival of the Internet which was unable to support KME rich Lisp-based environment.
See Also: