Antonym: The History Bots Edition

Minds are time machines.

There is no creature living I do fear and love so much as you.
– A letter to Thomas Cromwell

Dear Reader,

Some day soon we will scarcely remember a time when chatting with machines wasn’t as natural as conversing with one another.

All of this will turn out to be much stranger than it seems; but by the time we get there it will seem so ordinary that it will only be through simulations of ourselves as we were now that we will be able to understand the scale of the shift.

In Hilary Mantel's Wolf Hall trilogy, she did more than recount history—she crammed her mind so full of facts and data about the court of Henry VIII and the world around it that we might say she was running a simulation of it in her mind.

Her brain had a detailed model of Tudor England that she could ask questions of, set scenarios for characters and watch them resolve. This took immense focus and concentration. She had a small flat that she used only for writing, and there, surrounded by her notebooks and research, she said she slipped out of this world and into the one she was writing about. 

In a lecture called “I Met A Man Who Wasn’t There”, she said: 

Fiction provides a way to approach these gaps—airy spaces where a novelist can swing on a flying trapeze, rather than wade through muddy trenches of speculation as a biographer might.

Historical fiction authors create in those spaces between facts; places where Mantel discovered new truths, perspectives suggested so subtly by the sources that they have been missed. Patterns hiding in the data, insights clearly visible but not yet found:

In historical fiction, Cromwell has often been presented as either a bureaucratic drone or a villainous figure. Neither of these views aligns with contemporary evidence. He was clever, hardworking, and even his worst enemies acknowledged his talents. His closest friends both loved and feared him. One correspondent wrote to Cromwell, "There is no creature living I do love and fear so much as you."

Mantel’s fresh insights challenged accepted versions of the period. Eminent historians like Diarmud MacColluch reconsidered their views and embarked on new projects to explore them.

She had discovered new facts by making them up. Synthetic data, as we would call it if a generative AI large language model (LLM) had made (hallucinated?) them for us.

History bots

I expect somewhere there are historians using LLMs to make their own leaps. This week I read a paper (via Ethan Mollick) about an innovative approach for studying human behaviour using advanced technology known as Historical Large Language Models (HLLMs). These models are highly sophisticated systems capable of reading and interpreting historical texts.

Traditional research methods often focus on modern surveys and experiments, which overlook insights into how people in the past thought and behaved. The authors propose using HLLMs, trained on historical writings, to simulate how individuals from various historical periods and cultures might respond to questions about their attitudes and behaviours.

For instance, researchers could explore how Vikings, ancient Romans, or medieval Europeans might react to scenarios such as competitive games or debates on gender roles. This approach could provide valuable insights into human nature and its evolution over time. However, there are challenges to consider, such as the availability of historical texts for training these models and ensuring a fair representation of diverse social groups, not just the literate elite.

Despite the challenges, it will be fascinating to see how this method offers a glimpse into the thoughts and actions of our ancestors, broadening our understanding of human behaviour throughout history. Of course, we’ll never know for sure what they thought – but like Mantel’s Wolf Hall, the simulation might make us aware of other ways of seeing and thinking.

Better Call Cromwell

Aptly, one of the first AI bots I made a couple of years ago was a virtual Thomas Cromwell on Character.AI. For a while he was the unofficial agony aunt of Antonym, answering reader questions on matters of business and politics.

I checked in recently and you can now call him up. For some strange reason, he has an American accent at the moment. If I get some time, I will fix that.

Me, talking about AI literacy

If you happen to be a member of Agency Hackers, a group for mid-sized independent marketing services agencies, I’m giving a talk on AI literacy this Thursday (October 24th). You can find details here.

The full briefing on our AI literacy paper from two weeks ago is still up on YouTube. Early release versions of the paper have gone out to those who’ve asked for one. If you’re missing yours please do reply to me and I’ll send it straight away (a badly-timed switch of CRM systems last week means we may have missed you).

Here’s a short clip from the briefing of me explaining our definition of AI literacy:

Recommendations

Watching…

Ludwig (BBC iPlayer): Gentle. Warm. Murder. The classic Sunday night British detective show, but you can binge it anytime. A introverted genius who sets puzzles for a living has to – has to – impersonate his identical twin who is a senior detective in the picturesque city of Cambridge. I would have ignored this but for a lot of strong recommendations, and it was just what the doctor ordered, going down nicely with a cup of tea, thank you, vicar. David Mitchell does a great David Mitchell variant and it is wonderful.

Dune 2 (Various): The first time watching either of Denis Villeneuve’s Dune films at the cinema is rush. After the hype and hullaballoo has died down you get to revisit films like this that will have long half-lives as cult classics, imagination frames, and reference points as they take up residence in thousands of minds as an emotional reality. My second viewing was an absolute treat.

IC814 (Netflix): A shockingly realistic dramatisation of the 1999 hijacking of an Indian airliner. The most realistic feeling hijack drama I’ve seen (and there’s been a few lately hasn’t there?).

Reading…

Character Limit, by Kate Conger and Ryan Mac: Two New York Times reporters give a forensic account of Elon Musk’s takeover of Twitter. I read Walter Isaacson’s biography of Musk last year, which was interesting about Musk’s early life, but by the point Twitter was taken over the author seemed in thrall to his subject.

The Centre, by Ayesha Manazir Siddqi: This is fun. Anisa Ellahi, a Pakistani translator in London, has a boyfriend who is annoyingly good at speaking multiple languages. He eventually shares his secret with her: an intense, expensive ten day course that can make you fluent in any language. But what is really going on at…. THE CENTRE?

That’s all for this week…

Thank you for reading. I hope you found something interesting. If you did, please like, share or leave me a comment.

Antony