The British Broadcasting Corporation recently pulled off a stunt that felt straight out of a dystopian satire. On its flagship political debate show, Question Time, the producers decided to skip the usual panel of bickering politicians and pundits. Instead, they slapped digital avatars of Winston Churchill, Frida Kahlo, and Mahatma Gandhi onto screens to debate modern British politics.
It was bizarre. It was jarring. Honestly, it was a complete trainwreck.
The BBC pitched this as a tech demonstration, a way to spark public debate about the capabilities of generative artificial intelligence. What it actually did was highlight the sheer vacuity of treating complex historical legacies as interactive party tricks. When you program a machine to speak for a dead icon, you aren't resurrecting their intellect. You're just running a glorified autocomplete dressed up in digital skin.
The Illusion of Historical Authenticity
The primary flaw in the Question Time experiment lies in how these large language models operate. They don't think. They predict.
When the digital Winston Churchill weighed in on the UK's current economic struggles, the underlying software didn't analyze modern monetary policy through the lens of a 1940s wartime leader. It scraped a dataset of Churchill’s actual speeches, combined it with biographical data, and generated text based on statistical probabilities. The result is a caricature. It gives the audience what a computer thinks Churchill sounds like, relying on rhetorical flourishes, grand gestures, and predictable tropes.
This creates a massive authenticity gap. Take Frida Kahlo. She was a committed communist, an artist whose work was deeply intertwined with Mexican identity, anti-colonialism, and personal physical agony. Forcing her digital ghost to comment on British public transport infrastructure or contemporary European trade agreements is nonsensical. It strips away the nuance of her actual life and reduces her to an aesthetic.
The technology behind this relies heavily on deepfake audio synthesis and real-time animation. Companies like Synthesia and HeyGen have made massive strides in creating lifelike digital humans. The software maps facial movements to generated audio, matching lip sync and expressions with impressive accuracy.
Yet, the illusion shatters the moment the avatar speaks. The tone is frequently flat. The cadence is mechanical. The profound moral weight of someone like Mahatma Gandhi talking about peace becomes a series of generic platitudes when generated by an algorithm designed to avoid controversy.
Who Owns a Dead Icon's Voice
This BBC broadcast brings a massive legal and ethical quagmire into the mainstream spotlight. Who gave permission for these specific interpretations?
Living actors have spent the last few years striking and fighting for protection against digital cloning. The Screen Actors Guild-American Federation of Television and Radio Artists fought bitter battles over AI likeness rights. But the dead have very few protections.
In the United States, right of publicity laws vary wildly by state. California protects a celebrity's likeness for 70 years after their death, mostly to protect estate earnings. In the UK, where Question Time is produced, the legal framework is even murkier. There is no overarching "right of publicity" in English law. Instead, estates have to rely on trademark law or passing-off claims to protect a legacy.
When a public broadcaster uses the likeness of international historical figures, it opens a ethical minefield.
- The Problem of Bias: The data used to train these models reflects the biases of the internet. A digital Gandhi or Churchill will inherently lean toward the most common perspectives found online, erasing the complex, often controversial aspects of their true histories.
- The Issue of Consent: Dead figures cannot consent to their image being used to validate a specific modern political stance.
- The Risk of Weaponization: If a mainstream broadcaster can do this for entertainment, fringe political groups can do it to manufacture historical endorsement for radical ideologies.
Historians have expressed deep concern over this trend. Dr. Joan Tumblety, a cultural historian, has frequently pointed out that history is a process of constant re-interpretation based on evidence, not a static script to be fed into a chatbot. When we replace historical inquiry with automated simulation, we risk flattening our understanding of the past.
The Technological Shell Game
We need to talk about what this television experiment actually was. It wasn't an educational triumph. It was a gimmick designed to capture social media engagement.
The television production team used a combination of pre-rendered responses and real-time large language model querying. When an audience member asked a question, a team of producers monitored the inputs, fed them into the software, and selected the outputs that made for the best television. This isn't a live, autonomous debate. It’s a highly curated puppet show.
The real danger here is the erosion of truth. We live in an information ecosystem already plagued by synthetic media. Deepfakes of politicians are regular occurrences. By normalizing the creation of fully autonomous, talking AI versions of real people on trusted news platforms, the media trains audiences to accept synthetic reality.
If you can watch Churchill debate on television today, how will you verify a video of a modern leader saying something scandalous tomorrow? The boundary between reality and fabrication becomes dangerously blurred.
How to Interact with Historical AI Safely
The desire to use tech to engage with history isn't inherently evil. Museums and educational institutions have experimented with this for a while. The difference lies in intent and guardrails.
If you want to use generative technology to explore history without falling for the gimmicks, change your approach.
Stop treating chatbots as the actual historical figure. Don't prompt a tool with "Act like Winston Churchill and talk to me." Instead, ask the system to analyze historical documents. Use prompts like: "Based on the official letters written by Winston Churchill in 1940, what were his primary concerns regarding domestic morale?" This shifts the AI from a deceptive roleplay engine to a research assistant.
Verify the sources. If an app claims to give you a direct quote from a historical figure, copy that text and search for it in verified academic databases like Google Scholar or the digital archives of the British Library. If the quote only exists on blogs or social media, the software likely hallucinated it.
Demand transparency from media companies. When networks use synthetic media, they must clearly display a permanent watermark on the screen indicating the content is entirely artificial. Public broadcasters should be held to an even higher standard, ensuring that any use of digital avatars is accompanied by an explanation of the training data and software limitations.
We don't need digital resurrections to understand the past. We have their writings, their speeches, and their actual work. Turning giants of history into animated talking heads on a panel show doesn't make us smarter. It just makes our media environment a little more surreal, and a lot less trustworthy. Stick to the actual archives. The real people were far more interesting than any algorithm could ever simulate.
