The acoustics foyer at Bell Labs’ Murray Hill, N.J. research center in 1942.
Universal History Archive/Universal Images Group/Getty Images
In its heyday, AT&T’s Bell Labs was the center of innovation, akin to Silicon Valley today. With AI chatbots in the news, I wondered what happened to a now-vanished early version I used at the famed research institution’s New Jersey offices.
By Amy Feldman, Forbes Staff
I
was a Bell Labs kid. Like many who grew up in Berkeley Heights, New Jersey, down the road from the illustrious research institution’s Murray Hill headquarters, I had a father who worked at Bell Labs as an engineer.
In the mid-1970s, when I didn’t have school, my dad would sometimes take me to his office and I would play on Red Father, an early chatbot. In a room filled with hulking mainframe computers, I would sit at the keyboard and type to Red Father — the meaning of its name is lost to history, but perhaps alluding to the Cold War — and the machine would respond by text. Compared to board games like Monopoly or Battleship, playing with Red Father felt like being accepted into a secret society, a special game that only those of us who made it into the inner sanctum of the Labs’ leafy campus could use. The goal, in my mind, was to keep the conversation going as long as possible before Red Father, annoyed, would type back, “Go talk to your mother.”
With ChatGPT going viral, I was brought back to those days and wondered what had become of Red Father. It turns out that in the history of chatbots, Red Father exists only in the memories of a few people who played it. Neither AT&T’s corporate historian nor long-retired employees of Bell Labs knew about it, and after many calls failed to turn up information, I began to feel like I was chasing a ghost. Given Bell Labs’ history as an innovation center with researchers always playing around with new technologies, it’s likely that it was someone’s passion project, perhaps built solely for the fun of it, that never came close to having a commercial life.
Peter Bosch, now 61, recalls how when he was 14 his Bell Labs dad would bring the hardware with him from work so that he could play with it. “I used to love when he brought it home,” says Bosch, who spent his career as a software engineer. His goal, unlike mine, was to get Red Father annoyed as quickly as possible. “Your game was to draw it out and our game was to get to it as quickly as possible to annoy him,” Bosch says.
My dad passed three years ago at age 91, so I’m not able to ask him about Red Father. Among his circle of friends from those days who are still around, no one knows. Whoever developed the program would be quite elderly by now, if even still alive.
AT&T corporate historian Sheldon Hochheiser, who has been in that role since 1988, scoured the corporate archives and came up empty. “I can only speculate, but it would not be unusual for the researchers at Bell Labs to have such projects,” Hoccheiser says.
Bell Labs’ Claude Shannon, who did some of the earliest research in machine learning, demonstrates how his electronic mouse could learn its way around a maze without a mistake after only one ‘training’ run in 1952.
Keystone/Getty Images
Today, Silicon Valley is considered a hotbed of innovation, but in its heyday AT&T’s Bell Labs research facility was a center of technological research. William Shockley and two teammates invented the transistor there in 1947 and won a Nobel Prize. Two decades later, in 1969, Bell Labs’ researchers invented the Unix operating system. At its peak in the late-1960s, Bell Labs employed about 15,000 people, including 1,200 PhDs, as journalist Jon Gertner recounts in The Idea Factory: Bell Labs and the Great Age of American Innovation. “In a time before Google, the Labs sufficed as the country’s intellectual utopia,” Gertner writes.
Within that intellectual utopia, Bell Labs’ Claude Shannon, best known for establishing the field of information theory, did some of the earliest research in machine learning. In an early-1950s film demonstration, he showed how a life-sized magnetic mouse named Theseus navigated its way around a maze, remembering the directions that worked for future efforts. “He can learn from experience,” Shannon says in the film. “He can add new information and adapt to changes.”
Though Shannon’s work helped kick-start machine learning and pave the way for AI, Hochheiser, the AT&T historian, says that in Bell Labs’ archives the word “artificial intelligence” doesn’t show up in the titles of any technical memoranda until the 1980s. “I haven’t been able to really find much to answer the question of what happened between Shannon and the 1980s,” Hochheiser says. “If you look at the overall history of AI, the problem is that to do anything with artificial intelligence you needed far greater computer power than the computers of that era.”
The history of chatbots dates to the 1960s at MIT. In 1966, MIT computer scientist Joseph Weizenbaum developed Eliza, naming it after Eliza Doolittle in “My Fair Lady.”
“The Eliza program simulated a conversation between a patient and a psychotherapist by using a person’s responses to shape the computer’s replies,” according to MIT’s obituary of Weizenbaum. Though Eliza’s ability to communicate was limited, students and others who used it became drawn to it, sometimes revealing intimate details of their lives. While Eliza became a source of inspiration for other early chatbots, Weizenbaum became disillusioned with AI and later in his life cautioned against the technological advances that he’d once developed. In his 1976 book, Computer Power and Human Reason: From Judgment to Calculation, he warned about the potential dehumanization of computerized decision-making.
“Joe was very disconcerted by the reaction to Eliza, and he became a critic of AI optimism,” says Dave Clark, a senior research scientist at the MIT Computer Science and Artificial Intelligence Laboratory, who knew Weizenbaum. Eliza was originally written in a computer programming language that Weizenbaum had developed known as SLIP, and Clark says he is “willing to bet” that Weizenbaum developed Eliza to showcase the language. “He wanted to show what he could do with it,” Clark says. “And then he got spooked.”
Bell Labs’ researchers (left to right) John Bardeen, William Shockley and Walter Brattain won the Nobel Prize in 1956 for their invention of the transistor.
Hulton Archive/Getty Images
Bell Labs’ Red Father operated very similarly to Eliza, and was perhaps modeled on it. “It would try to parse as much information out of what you had entered, and use that to respond to you,” Bosch says. “It was an early attempt at a conversational interface with a computer. Very often it resorted to, ‘How does that make you feel?’ and ‘I’m sorry you don’t like bananas,’ or that sort of thing. A lot of the time it wasn’t that useful in terms of what it could pull out of your texts.”
Still, in the context of today’s buzz around chatbots, it’s bizarre and fascinating that there’s no record of it. “Often, like Red Father, those things aren’t well-documented,” Hochheiser says. “It’s clear when we’re looking back at the history of Bell Labs that researchers were given a lot of leeway in what they wanted to study.” As with Silicon Valley today, he says, researchers were often in their labs “whatever darn hours they felt like being there,” and brought in things they’d built at home.
A. Michael Noll, a professor emeritus at the University of Southern California who worked at Bell Labs in the 1960s and wrote a memoir about it, recalls that era of innovation. Researchers were working on all sorts of passion projects in the Labs’ heyday. He was working on digital computer art. “It was all the things you hear about in Silicon Valley today,” he says.
While Noll, 83, didn’t know anything about Red Father, he says it wouldn’t be surprising for someone, perhaps in the Unix area or in speech-processing, to have come up with it on the side. “A lot of stuff we did for fun,” he says. After all, he says, Bell Labs was part of AT&T and the parent company was more interested in a new telephone switching system than in computer art — or in an early chatbot that, to them, didn’t have obvious commercial applications. “People were looking into all these things that weren’t commercialized,” he says. “The list is probably a mile long. We had the freedom at Bell Labs to do weird things for a while.”
