The first chatbot was a joke. Only the oligarchs are laughing now

You have reached your maximum number of saved items.

Remove items from your saved list to add more.

I don’t know when I first heard the story of Joseph Weizenbaum, but it was a long time before our screens and news and thoughts and dreams and nightmares were so full of artificial intelligence.

He was on a temporary contract at the Massachusetts Institute of Technology in the mid-1960s, working for Project MAC, which had the aim of making computers more widely accessible. Seemingly to help achieve this aim, Weizenbaum developed the world’s first chatbot, on a computer the size of a room, that heated an entire building, and had just 150 kilobytes of memory.

Named after Eliza Doolittle from George Bernard Shaw’s Pygmalion, Weizenbaum’s creation was a play on Rogerian psychotherapy, a non-directive type of talk therapy created by Carl Rogers in the 1940s. Eliza could recognise words such as “mother” or “father”, change “I’s” to “you’s” and very, very basically, ask input sentences back as questions. If it couldn’t compute how to do this, it would use the disguise of Rogerian therapy to say things like, “please go on” or “I’m not sure I understand you fully.”

Weizenbaum actually had ulterior motives. He was already sceptical of the concept of AI, and saw Eliza as the chance to remind the world how farcical the concept of intelligent machines was. He wrote a paper about the program, publishing sections of its basic code, to say, “Look! This is so simple and yet it can appear intelligent!” Because, for him, appearance was as close as a computer could ever get to human intelligence.

Joseph Weizenbaum’s invention came to haunt him.Getty ImagesHe was sure his paper and this apparent joke would show the world how right he was. But the world didn’t take it the way he intended.

Eliza was celebrated. Prominent voices in the American therapeutic world thought it could be rolled out across the country to help with the therapy shortage. Then he discovered his own staff, the people who had helped create it, were sneaking in and using it without his knowledge and refusing to hand over their transcripts, calling them “private” conversations between “me and her”. Most famously of all, his secretary asked him to leave the room when she was “talking to Eliza”.

It has even come out that, in the very same building, under a disguise, was a government agency that also claimed the word “intelligence”; Weizenbaum, who had fled a growing Nazi Germany in 1936, was eventually forced to confront the fact the Pentagon was a major funder of his work.

His invention (or joke) had been received in exactly the opposite way he had intended. It was a nightmare for him. He received tenure at MIT on the back of this misunderstanding and nothing he could do seemed to sway people’s views of his creation.

I guess that’s the risk you take when you create and put something out into the world.

So, as the world was heating up over AI, and I wondered how I could creatively respond to all this, the first thought I had was to look back in order to understand something that seems to be so much about the future. I went to Weizenbaum, his secretary and their invention, and it was all there: all the contemporary worries about people falling in love with chatbots, about privacy, about environmental damage and, perhaps worst of all, all the legitimate fear about who is behind it and what their intentions are.

Dan Spielman as Joseph Weizenbaum and Manali Datar as his secretary, Becky, during rehearsals for Eliza.Charlie KinrossMy play is set in 1966, as Weizenbaum releases his creation and finds himself in a maelstrom of delusion.

To write it, I did something very un-AI: I talked to people. Weizenbaum’s daughters were a huge help. I talked to Jewish consultants about the early post-World War II Jewish experience, to computer engineers at MIT during the ’60s, to the people who rediscovered Weizenbaum’s files he had hidden in the MIT archives. I talked to friends and family and anyone and everyone to hear about their dreams and fears for this technology.

I also talked to a friend and collaborator of Weizenbaum about the mythical secretary who was never named, and then I got to share this name with an academic researching the lost women of computing history. Wonderfully, the academic and the secretary shared a name: Becky!

I also talked to Eliza; the program is freely available online thanks to members of the team who rediscovered the story. I even talked to ChatGPT and Claude and any other sentence-generating software to see how they would respond to prompts about Eliza and Weizenbaum. One of them tried to set the whole play in a jazz bar, with Weizenbaum wearing a beret.

What are we, as creatives, meant to think about this software? Is it just the latest in the long line of technology that has disrupted and brought fear to the arts, only to be used to create new, incredible work? Is it Baudelaire against the camera, or Bob Dylan coming out with an electric guitar?

AI oligarchs are the robber barons of our day. They steal our IP, churn through our water, tear apart our privacy, line their pockets in ways the Rockefellers could never have dreamed of, and threaten the structures of democracy.

As a citizen, all this enrages me. I dream of setting fire to military contracts, pumping polluted water into data centres and toppling space rockets as if they were statues of dictators. How dare these tech feudalist elites think they can steal everything we have ever created and earn a country’s GDP from the theft!

But there are rumblings of a different way: voices as diverse as Bernie Sanders and Steve Bannon have called for at least partial nationalisation. Cory Doctorow, coiner of the glorious term “enshittification” – an idea no AI could ever come up with – believes that when the financial bubble finally bursts, the hardware behind these models could be bought for pennies in the dollar and whatever is created from it can become open-sourced, as I believe it should have been in the first place. Imagine what creatives and scientists and all of us could do with unencumbered, collective human intelligence accessible through our fingertips, and free for all?

Because this is the rub with AI: The technology is glorious. Era-defining. Epoch-making. If we can get it out of the hands of the egomaniacs who run it, and use it to release, interrogate or deepen human creativity, then the possibilities are thrilling.

As I began to develop the play more, and watch the incredible team of creative people who are coming together to make it happen, I realised that theatre was exactly the right place to tell this story. The very nature of theatre, an act of communal human creativity being shared with other humans, becomes a form of protest in itself. By coming together to share something ephemeral, like performance, we are actively pushing against the pricks.

The play begins with a direct quote from Weizenbaum: “What I had not realised is that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people.”

I wonder if creating and going to the theatre, or music, or any type of communal art, is the very way to counter this delusion. We must gather, create, and reflect together.

Eliza is at Southbank Theatre, September 28-October 31.