Friday, May 26, 2023

Some History of Artificial Intelligence (AI) and Austin

MCC ca. 2005. Photo courtesy Larry D. Moore, CC BY 4.0 license

Note: this is an expanded version of an article originally written for Northwest Austin Civic Association (NWACA).

As the cliché goes, unless you've been living under a rock, you know Artificial Intelligence (AI) is all the rage. Major players like Microsoft and Google are investing big money in AI “chatbots” (AKA "stochastic parrots" by some researchers; see Bender 2021) like ChatGPT and Bard to enhance their search engines. In this article we step back in time and look at some history of AI in Austin.

While popular media makes the current round of chatbots synonymous with AI, AI has been around for years; the term “Artificial intelligence” was coined by computing pioneers in 1956 (Moor 2006). In the nearly 70 years since, AI research has involved more than just neural nets, the underpinnings of the current products like ChatGPT. What is considered AI is ever changing; a program that plays checkers certainly wouldn’t rate prime time news coverage these days but was once considered AI.

The history of AI is one of booms and busts. We are now in what AI researchers colloquially call the “third AI summer,” a boom time in AI research. The first two booms were followed by “AI winters” (the busts), periods when expectations exceeded results, or hardware evolution or economics failed to meet the demands of AI, leading to a decrease in interest, funding, and research.

The second big boom in AI might be said to have hit Austin in 1983 with the founding of the Microelectronics and Computer Technology Corporation (MCC). MCC was a consortium of companies helped along by the Department of Defense and the Justice Department as a response to Japan's Fifth Generation Computer Systems project, which was announced in 1982. Japan's Fifth Generation project utilized Prolog, a logic programming language popular in AI but second in popularity in the United States to the language Lisp. Japan’s Fifth Generation project was to AI like Russia’s Sputnik was to space exploration: a wake-up call that the United States was at risk of falling behind in a new technology. The iconic MCC building still stands today as the University of Texas West J.J. Pickle Research Building at the southwest corner of the intersection of MoPac and West Braker Lane.

While both the Fifth Generation project and MCC were causalities in part due to the second AI winter in the early 1990s, MCC is credited with helping jump-start Austin as a high-tech center. One of MCC’s spin-offs was a company called Cycorp which still does AI research headquartered in Austin. They’ve been working on AI for nearly four decades.

Cycorp’s product is Cyc, the name a play on “encyclopedia,” intending to create an encyclopedia of common-sense knowledge about the world. From their website, Cycorp is a “leading provider of Machine Reasoning AI. The Cyc platform combines an unparalleled common-sense ontology and knowledge base with a powerful reasoning engine and natural language interfaces...”.

Besides MCC and Cycorp, Austin has been home to a number of contributors to AI at the University of Texas and at companies like Schlumberger and Texas Instruments. Schlumberger’s Austin office is now closed, but the facility remains as Concordia University’s campus off FM 620. The Texas Instruments Explorer Lisp computer lost out to standard Unix machines in the second AI winter, victim to lower cost technology.

Will the current summer of AI be followed by a winter? Time will tell. Old AI never dies, it just morphs into other technology. Some feel the current neural net-based AI will eventually need to incorporate other techniques (“symbolic AI”) from the first two AI booms. If so, companies like Cycorp with roots going back to the 1980s may be well positioned to help. And the University of Texas graduate program in Computer Science, ranking #8 in the U.S. in AI, will ensure Austin stays in the AI game.

According to a U.S. News & World Report survey, in 2023 UT Graduate Computer Science ranked 8th out of 208 universities, tied with Georgia Tech and Princeton University. Four “specialties,” or areas of research, at UTCS also rank in the top ten, with Programming Languages coming in 7th, Artificial Intelligence ranked 8th, and Systems and Theory both at 9th.

Carton history of AI (Kautz, 2020). As this Tom and Jerry cartoon illustrates, neural nets weren't always the vanguard of AI. Click to enlarge.

References; Notes; Read More

Bender, Emily M. "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?" FAccT '21: Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, March 2021, Pages 610–623. Available on the ACM digital library, retrieved 5/27/2023

Bratko, Ivan. Prolog Programming for Artificial Intelligence. First published 1986.
Denney, Richard. Private conversation with self. What is "AI" changes over time. What is constant is the use of computers and computer science. The European Airbus (1980-1987) was the first commercial aircraft  to use computer-driven digital “fly-by-wire” (military use of fly-by-wire having preceded it) in which a computer calculates exactly which control surface adjustments are needed to make the aircraft respond to the pilot. Sounds a bit like the "generative AI" of tools like ChatGPT: the pilot says "bank left", the computer generates the needed changes in all control surfaces. Is that AI? Because "AI" is such a broad, vague category of techniques, we might do better dispensing with this term, helping demystify what is really happening. This would of course require the public, media and politicians to understand a little bit about how these "stochastic parrots" work. See Bender paper.
Feigenbaum, Edward and Pamela McCorduck. The Fifth Generation: Artificial Intelligence and Japan's Computer Challenge to the World. First published 1983. 

Kautz, Henry A. "The third AI summer: AAAI Robert S. Engelmore Memorial Lecture", Thirty-Fourth AAAI Conference on Artificial Intelligence, February 10, 2020.
Moor, James. The Dartmouth College Artificial Intelligence Conference: The Next Fifty Years. AI Magazine, Volume 27 Number 4, 2006 

No comments:

Post a Comment