Monday Morning Moan – what chatbots misunderstand about the history of AI

Innovations from World War II catalyzed significant advances in AI across several complementary approaches and disciplines. At the time, leading researchers worked together to combine logical reasoning, neural networks, and cybernetic techniques. But then a strange break occurred that effectively eliminated cybernetics from the field and use of the term.

A careful reading of first-hand sources revealed the likely cause of the dispute. Unsurprisingly, the chatbots did not. Getting to the heart of the matter required a careful consideration of the mostly unspoken motivations that can arise when the status quo is threatened and sexual norms are at risk of being violated. And this involved not only the imaginary sex parties, but also the ear piercings in the research teams.

Ultimately, the death of cybernetics may have more to do with the libertine love experiences of 1950s Chicago than the other things that were written down. Unfortunately, Large Language Models (LLMs) trained on people's writing don't capture these things because they're based on what people write and share. Of course, people write and share about far more controversial topics today than they did back then. However, many people may be reluctant to share controversial things in the workplace if doing so could lead to demotion, firing, or disempowerment. How much do you really want to talk to your employers or coworkers about drug use, narcissistic personality disorder, or ADD?

What happened to cybernetics?

In the 1990s I was fascinated by cybernetics and was heavily involved with Norbert Wiener, Stafford Beer and Howard & Eugen Odum who were exploring different ways of thinking about modeling and improving intelligence. When you talk about it today it's like a science fiction thing or a scientific discipline in the past tense. Of course, offshoots like cyberspace and cybersecurity (often shortened to cyber) are still used. And some of the core concepts are evolving in the area of ​​reinforcement learning and autonomous systems.

But cybernetics as a scientific discipline is dead. Perhaps the term has a few too many syllables. And it probably didn't help that Stafford Beer's work may have given power to a socialist regime in Chile that was later overthrown by a US-backed dictator fighting the Soviet threat. But none of this explained why cybernetics had suddenly disappeared.

I was recently reminded of the gap between cybernetics and the rest of AI when I read Neil D. Lawrence’s The atomic man. In a chapter on “gaslighting,” he alludes to a deliberate attempt to manipulate cybernetics pioneer Norbert Wiener, which suddenly triggered a very quick and decisive break with the rest of the AI ​​community at the time. It's a fascinating cautionary tale about how deliberate or random attempts to destroy trust can have an outsized impact on collaboration. But Lawrence never says who did the gaslighting and why. This seemed like a major oversight.

When I first read it, I imagined it was a colleague who was fed up with Wiener's blunt way of telling people what they were doing wrong. At the time of the break, Wiener had received numerous rejection letters for his memoirs, which said quite a lot of unflattering things about colleagues and family members. He also flattered the Soviets and complained about many potential problems we face today, such as unchecked automation and optimization of the wrong things.

Lawrence also suggests that Alan Turing may have accidentally been exposed to cyanide fumes from metal-plating experiments in his lab just below his bedroom. Most current literature suggests that this was a conscious response to the chemical castration he was sentenced to for being gay. But since little was written about homosexuality (when it was still a crime) and suicide (which was a sin) at the time, we will probably never know the truth about why this pioneering light of AI was extinguished so early in his career.

What the bots say

I'll get to what people have been saying about the sad divide between cybernetics and AI. But first, I recently decided to look at what the chatbots had picked up. I asked them all:

What can you tell me about the dispute between Norbert Wiener and Warren McCullough and this side of the AI ​​community in the early days of AI in the 1950s?

ChatGPT was pretty worthless and told me about the intense intellectual debates and disagreements, especially between Wiener and McCulloch, who used to be best friends and collaborators. There's a lot of talk about fundamental positioning, and ChatGPT throws in this nonsense:

The dispute between Wiener and McCulloch was not just a simple disagreement, but reflected deeper philosophical differences. Wiener was concerned that McCulloch's focus on replicating the human brain could lead to a dehumanizing development of AI. He feared that this could lead to systems that, while technically sophisticated, fail to take into account human values ​​and the potential consequences for society.

In response to further clarification on the subject of gaslighting, ChatGPT added that Wiener declined to attend the 1955 Dartmouth Conference, which shaped the future of AI, possibly due to a “fascinating aspect of potential misunderstanding.” Additionally, Wiener was led to believe that the famous Dartmouth Conference would contribute to his work.

Anthropic Claude summarized the rift between McCulloch and Wiener as being due to philosophical differences, different approaches to brain modeling, funding policies, and personal and professional rivalries. On the gaslighting issue, it was even less helpful. Google Gemini summarized the rift as being due to philosophical differences, different scientific emphases, and personality conflicts. Google Gemini also added this extremely unhelpful insight on the topic of gaslighting:

There is no comprehensively documented or credible evidence that Norbert Wiener was prevented from attending a conference due to any form of gaslighting or deception.

Google Gemini continued to argue health issues, conflicting commitments and personal reasons. It also added a definition of gaslighting and asked for clarification based on “extensive evidence.” It still didn't mention concerns about Chicago's permissive sex parties.

What people say

So what do people have to say about this? Well, according to the 2004 book Dark hero of the information age by Flo Conway and Jim Siegelman. The trigger for the cybernetic break was Wiener's wife Margaret, who was concerned about the influence McCulloch's promiscuous lifestyle might have on her husband and how it might negatively affect her social status within the MIT community.

This is all evident from the testimony of Wiener's two daughters and one of his closest confidants, who were furious at Margaret's behavior. Everyone else in this previously close-knit community was somewhat surprised by the rift. A more benevolent interpretation might have been that she was concerned about Norbert's mental well-being, although her daughters argued that, on the whole, she did not think so.

By this time, Margaret Wiener had already gotten an MIT employee fired for getting her ears pierced, and she was immediately disgusted by McCulloch's extravagance at their only dinner party together. And back then, people didn't talk much about explicit topics like sexual misconduct or ear piercings, even in private.

So this was the gaslighting incident, according to this report: Margaret told a mutual friend that her more flawless daughter had been deflowered multiple times by different people while visiting these promiscuous professors in Chicago. All things considered, this may have been entirely believable, but according to the daughters in question, it was also entirely fabricated.

The consequences were swift and decisive. The day after hearing this, Wiener wrote a nasty letter to his MIT colleagues, claiming that the McCulloch team could not be trusted with research funding and that they were doing poor work. And this after more than a decade of collaboration. Here is Conway & Siegelman's summary:

At his best, Wiener was exuberant, impulsive, and often moody. At his lowest points, he succumbed to crippling depressions that drove him to frequently threaten suicide in the circles of his home and family, and sometimes among his MIT colleagues. But in many ways, Wiener's extremes matched those of his wife, a demanding professor modeled after the Old World. In her dutiful efforts to protect her nervous husband, Margaret Wiener took steps to neutralize Wiener's colleagues, women around him, and anyone she viewed as a threat to his notoriety. One strategy in particular backfired disastrously for him personally and professionally.

For a decade, Wiener collaborated productively with the pioneering neuroscientist Warren McCulloch and Walter Pitts, the young genius of the next generation of cybernetics. The sudden end of his collaboration with McCulloch, Pitts, and other talented young scientists who came to MIT to advance cybernetics was a crisis for Wiener and everyone involved. This separation dealt a severe blow to the cybernetics revolution at a crucial moment and changed the course of the new technological era in ways that continue to resonate today.

My opinion

I was a little baffled by LLM's current inability to uncover the root cause of the AI ​​divide, nearly seventy years later and twenty years after the public debate. It seems reasonable to assume that differences of opinion can affect things profoundly, and humans have an amazing ability to read between the lines, even when most of the time people don't write or say.

But the disagreements we don't talk or write about don't always have to be about sensitive issues. They can also be simple differences between what managers say and do and how they affect employees. People need to think about these things to ensure their companies continue to be successful, and to understand what's going on with their suppliers and customers, who might also be inclined to twist the facts for a variety of reasons.