In the 60s, Weizenbaum tried to show the limitations of AI. No-one listened

In 1966, German American computer scientist Joseph Weizenbaum, a professor at the Massachusetts Institute of Technology (MIT), created a computer program he dubbed ELIZA.

It was a rudimentary version of artificial intelligence, named after Eliza Doolittle, the working-class woman from George Bernard Shaw's Pygmalion who learns how to behave like a "lady".

The program could do little more than repeat back what it was told, phrased as a question. For Weizenbaum, who escaped Nazi Germany 30 years earlier, ELIZA was a way to highlight the limitations of AI.

For some users, including his secretary, it became almost a companion; when she noticed Weizenbaum watching her use ELIZA, she said: "Would you mind leaving the room, please?"

Now, AI pervades our lives; some people use it as a therapist, researcher, confidante. But it has also stirred up complicated legal conversations, as AI companies attempt to override Australia's copyright laws, even as experts like Bill Gates warn of the technology's potentially lethal consequences.

That's not to mention AI's impact on the environment, with Sydney Water warning data centres could use the equivalent of a quarter of the city's current annual drinking water supply within a decade.

As these conversations continue, artists are also grappling with the impacts of AI. For playwright Tom Holloway, that meant going back to the story of Weizenbaum and ELIZA for his new play, also called Eliza, now at Melbourne Theatre Company.

I can't remember when I first heard the Weizenbaum story. It was like 10, 15 years ago. It might have been a Radiolab podcast or an Adam Curtis doco. But it just hit me.

There was something deeply human about these conversations between a human and this computer program. There was a deep need and loneliness in the original transcripts that had been shared.

The story is more complicated because Weizenbaum was an AI sceptic and created the program to say, look, this thing can seem intelligent, yet it's absolutely not. It's only 400 lines of basic code. He was trying to use this chatbot of ELIZA to highlight that a computer could never understand human intelligence and never really connect with a human, and it was interpreted in exactly the opposite way. Even that seemed deeply human.

The story stayed with me, and then a few years ago, when we all started living in this same world as Weizenbaum's nightmare, it seemed like the story that needed to be told. Because we forget, don't we? We don't know this has been something we've been dealing with for 60 years and much longer too.

Every new technology brings terror and causes disruption before we find ways to work with it.

What I love about stories like Weizenbaum's is that humanity, that sense of the person grappling with the thing they have made, like Mary Shelley's Frankenstein or, more recently, Christopher Nolan's Oppenheimer.

If we remember AI is simply a technology, the tool is incredible, and as long as we double-check everything it tells us, it can give us access to so much that humans have articulated for all time, because it's stolen that from us.

And that's the problem with it. The problem with it is how it's been created, the people behind it, and what it does to the environment. We just need to regulate it and work out how to stop it destroying the environment and free it for us all, reclaim it, because it's built on us.

It can be frustrating as a journalist to hear accusations about writers using "AI voice" or that em dashes are a tell for someone using AI, when AI was trained on our writing.

I had to sign a thing because some of my plays had been discovered to have been illegally used in the training of one of the models!

How did you feel when you found out AI had been trained on your work?

It might have been my first play, Beyond the Neck. It's about how humans and communities survive something as terrible as the Port Arthur massacre. It's this attempt to talk about something very difficult and deeply human.

I wasn't surprised because we already knew [stealing creative work] is what it has done. It continues to be shocking that we allowed that to happen, and not just that, but the levels [AI companies would] go to. There was the story that Anthropic [was] specifically trying to get rare books, feed them into the machines, and then destroy them, so that knowledge wasn't anywhere else.

That's more than just taking what's on the internet. That is specifically trying to control knowledge, and that's awful. That's the stuff that we should be rioting over.

Obviously, there are worrying things that happen with AI, but that's because of the people giving the prompts and not being clear enough, and there's not enough of a safety net around those things.

The Medicare breach was, from what I've read — as a playwright, I feel quite far out of my element talking about some of this stuff — it was asked to get some health statistics, and it did it by whatever means it needed to. And the Hugging Face thing was quite similar.

It all comes back to the human and human error, and a lack of a safety net built in by us humans. That's why I think regulation is key.

Has your relationship to or understanding of AI changed since you first started writing Eliza?

I've been writing it for about three years, so I think everyone's relationship to it has changed over that time. It's almost like an abusive relationship. Humans are using it every day, and yet we're also terrified of it. Was it earlier this year there were all those speeches given at graduation ceremonies in America where guests were talking about it as a positive thing, and the students all booed?

The thing that writing Eliza has really brought home is it's not the tech that I'm worried about; it's those people [behind it]. And that was a real revelation for Weizenbaum back then too. He said he was shocked at how quickly reasonably minded people were deluded by technology. And that's not just the people using it; that's the people making money from it too and using it to manipulate other people.

Yes, I use it, and I have all my settings set up so that it's not allowed to use any of my chats. I delete the chats as soon as they've happened. I donate to environmental charities to make up for it.

Because I think the actual technology can push you intellectually and creatively. If you relinquish those skills to it, there is danger there, and you are only going to get worse work back from it.

But it can stimulate you, push you, and make offers that can help advance the work or creativity you're doing yourself.

If we can free AI, I think there's a future where artists will use it as a tool to create things we can't even imagine now.

Weizenbaum got really disillusioned and hid a lot of the original coding and transcripts [for ELIZA] in the MIT archives. And there's this team of some of the biggest cybernetics brains in the world from Stanford and Oxford and other universities, who during COVID went back in and found the lost files. They've been reading drafts and giving me tidbits and facts and helping me.

In writing this play about a technology isolating people, I've grown more connections with more people, and discovered theatre is the right form to tell this story because it becomes a kind of act of protest against that isolation [AI companies] are hell-bent on forcing us all into.

The communal nature of the creative process in theatre, and then with audiences, feels like we're coming together to go, "No, we're what's important in all this."

So, would you like to see artists' copyright be protected?

To be an artist, you have to accept risk and discomfort into your daily life, and you do it because you want to try to say something that might help other people.

Whenever I see anyone else putting anything into the world, I think that person has risked so much. There's something very hopeful about the act and very generous about the act that I think needs protecting.

Eliza is at Melbourne Theatre Company until October 31.

This interview has been edited for clarity and brevity.