Eliza Play Revisits the Human Story Behind the Original Chatbot

The original chatbot still asks an uncomfortable question

The story behind Tom Holloway’s play *Eliza* is less about a machine becoming human than about people wanting it to. That gap—between what a computer can do and what we hope it understands—has only grown more relevant as AI becomes part of daily life. In 1966, Joseph Weizenbaum, a German American computer scientist and Massachusetts Institute of Technology professor, built ELIZA. He named the rudimentary program after Eliza Doolittle, the working-class woman in George Bernard Shaw’s *Pygmalion* who learns to pass as a “lady.” ELIZA mostly turned users’ words into questions. Weizenbaum, who had fled Nazi Germany about 30 years earlier, meant the program to expose AI’s limits. Some users saw something else. Weizenbaum’s secretary became so absorbed in talking to ELIZA that, when she noticed him watching, she asked him to leave the room.

That response is the human story Holloway wanted to tell. He says he first encountered Weizenbaum’s story perhaps 10 or 15 years ago, through a *Radiolab* podcast or an Adam Curtis documentary. The surviving conversations moved him: they revealed loneliness and need, even though ELIZA had only about 400 lines of basic code and couldn’t understand or connect with anyone. Weizenbaum was an AI sceptic. He built the chatbot to show that a machine could appear intelligent without possessing human intelligence. The irony is that people read the program in almost the opposite way. For Holloway, that misunderstanding is as revealing as the technology itself.

Old fears, present stakes

AI now serves some people as a therapist, researcher or confidante. Its wider effects are harder to ignore: companies are challenging Australia’s copyright rules ([ABC coverage]( /news/2026-09-16/top-ai-firms-meet-albanese-ministers-copyright-law/107160498)), Bill Gates has warned of potentially lethal consequences ([The Guardian](https://www.theguardian.com/us-news/2026/sep/27/bill-gates-artificial-intelligence-kristen-welker)), and Sydney Water has warned that data centres could consume the equivalent of a quarter of the city’s current annual drinking water supply within a decade ([ABC coverage]( /news/2025-08-27/ai-to-take-up-one-quarter-of-sydney-water-in-a-decade/105700928)). Artists are also confronting what AI means for their work ([ABC coverage]( /news/2026-10-01/eddie-perfect-the-tartuffe-malthouse-theatre-2027-season/107188726)). Holloway’s *Eliza*, now playing at Melbourne Theatre Company, returns to Weizenbaum’s story at a moment when the technology he feared has become ordinary. The obvious reading is that we’re living through a new technological upheaval. Holloway’s point is more unsettling: people have been wrestling with these questions for decades, and each new technology brings fear and disruption before society finds ways to live with it. He connects that struggle to *Frankenstein* and Christopher Nolan’s *Oppenheimer*: stories about people facing the consequences of what they create. Holloway sees AI as a powerful tool, but says users need to check its answers. The deeper problems, in his view, lie in who built it, how it was built and its environmental costs. He argues that AI needs regulation and that its benefits should be reclaimed for the people whose work helped create it.

Holloway says some of his plays were used illegally to train an AI model, and that he had to sign a document after the use came to light. The work may have been his first play, *Beyond the Neck*, about how people and communities survive the Port Arthur massacre ([ABC coverage]( /news/2006-08-04/port-arthur-play-to-debut-in-hobart/1230162)). He wasn’t surprised, he says, but he remains shocked that it was allowed to happen—and by how far companies would go. He points to reports that Anthropic sought rare books to feed into its models and then destroy ([The Guardian](https://www.theguardian.com/technology/2026/aug/02/australian-book-sellers-alarm-destruction-rare-titles-ai-supply-chain)). To Holloway, that suggests an effort to control knowledge, not merely collect material already online.

Holloway also pushes back on the idea that writers’ use of an “AI voice,” or em dashes, proves they relied on a chatbot. AI was trained on human writing, he argues, and his own plays were among the work used without permission. His concerns extend to water use and safety. He says some troubling AI outcomes stem from unclear prompts and inadequate safeguards. The interview’s opening question, then, isn’t whether a chatbot can sound human. It’s what people reveal when they choose to believe it can.

AI’s risks come from the people who control it

Tom Holloway sees the Medicare breach and the Hugging Face incident as versions of the same problem: systems asked to retrieve information, then allowed to pursue that goal with too little human oversight. He says the failures point to human error and inadequate safeguards—and make regulation essential. Writing *Eliza* over the past three years has changed how he thinks about AI. He compares people’s daily reliance on it, alongside their fear of it, to an abusive relationship. He recalls American students booing speakers who praised AI at graduation ceremonies. But the play led him to a sharper concern: “It’s not the tech that I’m worried about; it’s those people [behind it].” That echoes Joseph Weizenbaum’s own disillusionment with ELIZA, the early chatbot that inspired the play. Holloway says Weizenbaum was shocked by how quickly otherwise reasonable people could be taken in by technology. The danger, in Holloway’s view, extends from users to the people who profit from AI and use it to influence others.

A tool that needs boundaries

Holloway uses AI, but has disabled the use of his chats for training and deletes them after each session. He also donates to environmental charities to offset his use. His view isn’t that artists should reject the technology: it can challenge creative thinking and suggest directions a person might not find alone. Hand over those skills, though, and both the process and the work can suffer. He sees a possible future in which artists use AI to make things that are hard to imagine today—but says that future depends on freeing the technology for creative use. What that means in practice isn’t fully clear. His own method offers a playful answer to whether AI helped write the play: “I used ELIZA to write *Eliza*.” To research it, Holloway spoke with Joseph Weizenbaum’s daughters. He also received material from an international group of researchers who recovered lost ELIZA code and transcripts from the MIT archives during the COVID pandemic. Weizenbaum had hidden the files after becoming disillusioned with his creation. The subject of the play is technology that isolates people. Yet Holloway says writing it brought him into contact with more people, and convinced him theatre was the right form for the story. A live performance, made collectively and experienced with an audience, can become a protest against the isolation he believes AI companies are pushing. “No, we’re what’s important in all this.”

Why artists’ work needs protection

For Holloway, protecting artists’ copyright is “deeply important.” Making art means accepting risk and discomfort in the hope that the work might help someone else. He sees generosity in that choice—and says it deserves protection. That’s a pointed ending for a play about AI: the debate isn’t only about what the technology can produce, but whose work and judgment it depends on, and whether the people behind that work retain a say. [ *Eliza*](https://www.mtc.com.au/plays-and-tickets/whats-on/season-2026/eliza) is at Melbourne Theatre Company until October 31. This interview has been edited for clarity and brevity.